Apparatuses, systems and methods for VIAL seal inspection using silhouette images

Silhouette imaging with telecentrically arranged cameras and multiple angled cameras addresses false rejects and accepts in vial seal crimp inspection by providing clear image delineation and multi-metric analysis, ensuring accurate and efficient defect detection.

WO2026236378A1PCT designated stage Publication Date: 2026-11-12AMGEN INC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/US2026/026925
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-07
Filing Date
2026-05-06
Publication Date
2026-11-12

AI Technical Summary

Technical Problem

Two-dimensional vision-based vial seal crimp inspection systems suffer from high rates of false rejects and false accepts due to illumination conditions, lensing artifacts, and variations in crimp material properties, particularly in anodized aluminum seals, which affect reflectivity and color, leading to inconsistent defect detection outcomes.

Method used

The use of silhouette imaging techniques with telecentrically arranged cameras and backlights generates high-contrast images with well-defined edges, allowing for clear delineation between the vial seal crimp and background, and employs multiple cameras at angles to provide comprehensive coverage without vial rotation, using multi-metric alert conditions to distinguish between defective and non-defective seal crimps.

Benefits of technology

This approach reduces sensitivity to seal surface reflectivity and color variations, maintains high throughput inspection, and accurately detects seal crimp defects with reduced mechanical complexity, enhancing the reliability of defect detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure US2026026925_12112026_PF_FP_ABST
    Figure US2026026925_12112026_PF_FP_ABST
Patent Text Reader

Abstract

Method for vial seal inspection including generating a silhouette image of a vial located between a camera and a backlight at a defect detection station of an assembly line system, the vial having an associated crimp type; identifying a plurality of zones within the silhouette image, wherein the plurality of zones are defined based on a reference pattern associated with the crimp type; analyzing the silhouette image to measure a plurality of evaluation metrics for a seal crimp of the vial, wherein the measurements are based on image data within the identified plurality of zones; determining whether the seal crimp of the vial includes a defect by determining whether at least one of the plurality of evaluation metrics exhibits a corresponding alert condition; and rejecting or accepting the vial based on whether the seal crimp of the vial includes the defect.
Need to check novelty before this filing date? Find Prior Art

Description

Electronically Filed May 6, 2026APPARATUSES, SYSTEMS AND METHODS FOR VIAL SEAL INSPECTION USING SILHOUETTE IMAGESCROSS REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Patent Application No.63,801,363 filed May 7, 2025.FIELD OF DISCLOSURE

[0002] The present disclosure generally relates to apparatuses, systems, and methods for vial seal inspection. More particularly, the present disclosure relates to apparatuses, systems, and methods for vial seal inspection using a defect detecting system that generates and analyzes silhouette images of a vial.BACKGROUND

[0003] Under FDA requirements, many products (e.g., filled and sealed vials of medicine, etc.) need to be inspected for defects. Vial seal crimp inspection is one example type of defect inspection. Vial seal crimp inspection is traditionally implemented using two-dimensional (2D) computer vision-inspection. Among other sources of error, known 2D vial seal crimp inspection is prone to errors due to associated illumination conditions, and lensing artifacts that limit such imaging systems from producing clearly delineated boundaries between a seal crimp of a vial and background portion of an image.

[0004] Another source of error is caused by variations in crimp material (e.g, an anodized form of aluminum, etc.) resulting in different levels of reflections. Therefore, when an associated 2D vision system is configured to, for example, detect defects on one batch of aluminum vial seals, the system may achieve success defect inspection on a first batch of sealed vials but not other batches of sealed due to variations in seal surface finishing, slight variations in vial seal surface color, etc. Moreover, because traditionally 2D vial seal crimp inspection techniques often include illuminating an associated vial seal crimp area during inspection, differences in reflectivity across batches may introduce differences in the defect detection outcomes. All these issues may cause false rejects (e.g, a sealed vial having a good vial seal crimp erroneously identified as being a bad vial seal) or false accepts (e.g., a sealed vial having a improper vial seal crimp erroneously identified as being a good vial seal). In any event, known inspection system errors often result in false rejects and / or false accepts.Electronically Filed May 6, 2026

[0005] Additionally, vial seal inspection systems are configured to operate at high throughput rates (e.g., 600 or more parts per minute), which limits the time available for image acquisition and analysis and further constrains the complexity of inspection algorithms that can be employed.

[0006] A vial seal crimping process often includes a mechanical process performed after successful filling of an associated vial. As described in detail herein, the mechanical nature of vial seal crimping processes and / or variations in associated vials (e.g., a glass container, etc.) often results in vial seal crimp defects.

[0007] As part of a vial inspection, known 2D vial seal crimp inspection may be incorporated into an automatic visual inspection (AVI) system. In known vial seal inspection systems, a sealed vial may be rotated 360 degrees about a central axis of the vial, multiple two-dimensional (2D) images may be taken at various points about the vial rotation, and these 2D images may be analyzed using, for example, machine vision-based algorithms to detect seal defects. False rejects or false accepts often occur due to the above-described variations of associated 2D vision-based systems.SUMMARY

[0008] The present disclosure addresses limitations of two-dimensional vision-based vial seal crimp inspection systems, including unacceptable rates of false rejects and false accepts due to illumination conditions, lensing artifacts, and variations in crimp material properties such as differences in reflectivity across batches of anodized aluminum seals. The disclosed apparatuses, systems, and methods utilize silhouette imaging techniques that generate high-contrast images with well-defined edges by telecentrically arranging cameras and backlights, thereby reducing sensitivity to seal surface reflectivity and color variations. The silhouette images can clearly delineate the boundary between the vial seal crimp and the background, enabling efficient and accurate measurement of evaluation metrics such as angles, areas, and distances within defined zones of the silhouette image. The use of zones defined based on a reference pattern reduces the amount of image processing used, thereby maintaining high throughput inspection. Furthermore, the use of multiple cameras oriented at angles relative to each other can provide improved coverage of the seal crimp area without requiring vial rotation during inspection, thereby increasing throughput while maintaining comprehensive defect detection capability and reducing mechanical complexity. The disclosed multi-metric alert conditions can enable fine distinction between defective and non-defective seal crimps, particularly in ambiguous boundary cases where individual metrics alone may be insufficient. Additionally, because the disclosed multi-metric alert conditions can include simple BooleanElectronically Filed May 6, 2026 expressions, they can efficiently determine whether a seal crimp includes a defect while maintaining the high throughput needed for crimp inspection.

[0009] In some aspects, the techniques described herein relate to a method including: generating a silhouette image of a vial located between a camera and a backlight at a defect detection station of an assembly line system, the vial having an associated crimp type; identifying, via one or more processors, a plurality of zones within the silhouette image, wherein the plurality of zones are defined based on a reference pattern associated with the crimp type; analyzing, via the one or more processors, the silhouette image to measure a plurality of evaluation metrics for a seal crimp of the vial, wherein the measurements are based on image data within the identified plurality of zones; determining, via the one or more processors, whether the seal crimp of the vial includes a defect by determining whether at least one of the plurality of evaluation metrics exhibits a corresponding alert condition; and rejecting or accepting the vial based on whether the seal crimp of the vial includes the defect.

[0010] In some aspects, the techniques described herein relate to a defect detection system including: at least one camera; at least one backlight arranged telecentrically with the at least one camera; a memory storing a reference pattern associated with a vial crimp type; and a control unit having one or more processors operatively connected with the at least one camera and the memory and one or more non-transitory, computer-readable media storing instructions that, when executed by the one or more processors, cause the control unit to: receive a silhouette image of a vial generated by the at least one camera while the vial is located between the at least one camera and the at least one backlight, the vial having the vial crimp type; identify a plurality of zones within the silhouette image, wherein the plurality of zones are defined based on a reference pattern associated with the crimp type; analyze the silhouette image to measure a plurality of evaluation metrics for a seal crimp of the vial, wherein the measurements are based on image data within the identified plurality of zones; determine that the seal crimp of the vial includes a defect by determining that at least one of the plurality of evaluation metrics exhibits a corresponding alert condition; and generate an alert signal indicating that the vial is defective based on whether the seal crimp of the vial includes the defect.

[0011] In some aspects, the techniques described herein relate to a method including: generating a silhouette image of a vial located between a camera and a backlight at a defect detection station of an assembly line system, the vial having an associated crimp type; determining, via one or more processors, whether the silhouette image matches a reference pattern associated with the crimp type, the reference pattern defining a plurality of zones toElectronically Filed May 6, 2026 be located within the silhouette image; determining, via the one or more processors, whether the seal crimp of the vial includes a defect by determining whether the silhouette image fails to match the reference pattern associated with the crimp type; and rejecting or accepting the vial based on whether the seal crimp of the vial includes the defect.

[0012] In some aspects, the techniques described herein relate to a defect detection system including: at least one camera; at least one backlight arranged telecentrically with the at least one camera; a memory storing a reference pattern associated with a vial crimp type; and a control unit having one or more processors operatively connected with the at least one camera and the memory and one or more non-transitory, computer-readable media storing instructions that, when executed by the one or more processors, cause the control unit to: receive a silhouette image of a vial generated by the at least one camera while the vial is located between the at least one camera and the at least one backlight, the vial having the vial crimp type; determine whether the silhouette image matches a reference pattern associated with the crimp type, the reference pattern defining a plurality of zones to be located within the silhouette image; determine whether the seal crimp of the vial includes a defect by determining whether the silhouette image fails to match the reference pattern associated with the crimp type; and generate an alert signal indicating that the vial is defective based on whether the seal crimp of the vial includes the defect.

[0013] In some aspects, the techniques described herein relate to a method including: generating a silhouette image of a vial located between a camera and a backlight at a defect detection station of an assembly line system, the vial having an associated crimp type; identifying, via one or more processors, a plurality of zones within the silhouette image, wherein the plurality of zones are defined based on a reference pattern associated with the crimp type; analyzing, via the one or more processors, the silhouette image to measure a plurality of evaluation metrics for a seal crimp of the vial, wherein the measurements are based on image data within the identified plurality of zones; determining, via the one or more processors, whether the seal crimp of the vial includes a defect by determining whether at least one of the plurality of evaluation metrics exhibits a corresponding alert condition; and rejecting or accepting the vial by determining whether the seal crimp of the vial includes the defect.

[0014] In some aspects, the techniques described herein relate to a defect detection system including: at least one camera; at least one backlight arranged telecentrically with the at least one camera; a memory storing a reference pattern associated with a vial crimp type; and a control unit having one or more processors operatively connected with the at least oneElectronically Filed May 6, 2026 camera and the memory and one or more non-transitory, computer-readable media storing instructions that, when executed by the one or more processors, cause the control unit to: receive a silhouette image of a vial generated by the at least one camera while the vial is located between the at least one camera and the at least one backlight, the vial having the vial crimp type; identify a plurality of zones within the silhouette image, wherein the plurality of zones are defined based on a reference pattern associated with the crimp type; analyze the silhouette image to measure a plurality of evaluation metrics for a seal crimp of the vial, wherein the measurements are based on image data within the identified plurality of zones; determine that the seal crimp of the vial includes a defect by determining that at least one of the plurality of evaluation metrics exhibits a corresponding alert condition; and generate an alert signal indicating that the vial is defective by determining that the seal crimp of the vial includes the defect.

[0015] In some aspects, the techniques described herein relate to a method including: generating a silhouette image of a vial located between a camera and a backlight at a defect detection station of an assembly line system, the vial having an associated crimp type; determining, via the one or more processors, whether the silhouette image matches a reference pattern associated with the crimp type, the reference pattern defining a plurality of zones to be located within the silhouette image; determining, via the one or more processors, whether the seal crimp of the vial includes a defect by determining whether the silhouette image fails to match the reference pattern associated with the crimp type; and rejecting or accepting the vial by determining whether the seal crimp of the vial includes the defect.

[0016] In some aspects, the techniques described herein relate to a defect detection system including: at least one camera; at least one backlight arranged telecentrically with the at least one camera; a memory storing a reference pattern associated with a vial crimp type; and a control unit having one or more processors operatively connected with the at least one camera and the memory and one or more non-transitory, computer-readable media storing instructions that, when executed by the one or more processors, cause the control unit to: receive a silhouette image of a vial generated by the at least one camera while the vial is located between the at least one camera and the at least one backlight, the vial having the vial crimp type; determine whether the silhouette image matches a reference pattern associated with the crimp type, the reference pattern defining a plurality of zones to be located within the silhouette image; determine whether the seal crimp of the vial includes a defect by determining whether the silhouette image fails to match the reference pattern associated withElectronically Filed May 6, 2026 the crimp type; and generate an alert signal indicating that the vial is defective by determining that the seal crimp of the vial includes the defect.BRIEF DESCRIPTION OF THE DRAWINGS

[0017] It is believed that the disclosure will be more fully understood from the following description taken in conjunction with the accompanying drawings. Some of the drawings may have been simplified by the omission of selected elements for the purpose of more clearly showing other elements. Such omissions of elements in some drawings are not necessarily indicated of the presence or absence of particular elements in any of the exemplary embodiments, except as may be explicitly delineated in the corresponding written description. Also, none of the drawings are necessarily to scale.

[0018] FIG. 1 depicts an example sealed vial having a flip cap, according to some embodiments;

[0019] FIGs. 2A-2C depict example components for sealing a vial, according to some embodiments;

[0020] FIG. 3 A depicts an example system for crimping a vial seal using high speed rotating blades, according to some embodiments;

[0021] FIG. 3B depicts a chart of representative acceptable and unacceptable crimp types, according to some embodiments;

[0022] FIGs. 4A and 4B depict a defect detection station of an assembly line system that includes an example vial seal inspection system, according to some embodiments;

[0023] FIG. 4C depicts a representation of a field of view for cameras of the example vial seal inspection system shown in FIGs 4A and 4B, according to some embodiments;

[0024] FIG. 4D depicts another example vial seal inspection system, according to some embodiment;

[0025] FIG. 4E depicts a representation of a field of view for cameras of the example vial seal inspection system shown in FIG. 4D, according to some embodiments;

[0026] FIG. 5 A depicts a silhouette image of a seal crimp for a vial captured using the example vial seal inspection systems of FIGs. 4 A or 4D, according to some embodiments;

[0027] FIG. 5B depicts an example reference pattern of zones in relation to the silhouette image of FIG. 5 A, according to some embodiments;

[0028] FIGs. 5C-5M depict example reference locations identified within the zones depicted in Fig. 5B, according to some embodiments;

[0029] FIGs. 6A-6Q depict example evaluation metrics measured with respect to the reference locations depicted in FIGs. 5C-5M, according to some embodiments;Electronically Filed May 6, 2026

[0030] FIG. 7A depicts a silhouette image of a seal crimp for another example vial captured using the example vial seal inspection systems of FIGs. 4A or 4D, according to some embodiments;

[0031] FIG. 7B depicts example evaluation metrics measured with respect to identified reference locations of the silhouette image of Fig. 7A, according to some embodiments;

[0032] FIGs. 8A-8C depicts example silhouette images with defective seal crimps determined by failure to match a defined reference pattern, according to some embodiments;

[0033] FIG. 9A depicts a flow chart for a method of generating absolute alert thresholds for one or more evaluation metrics, according to some embodiments;

[0034] FIG. 9B depicts a flow chart for a method of generating components of a multimetric alert condition for evaluating evaluation metrics, according to some embodiments;

[0035] FIGs. 10 A- 10C depict graphs of evaluation metrics calculated for defective and non-defective vial sample for use in determining threshold values used to indicate a defective seal crimp; and

[0036] FIG. 11 depicts a flow chart of a method for identifying a defective seal crimp of a vial, according to some embodiments.

[0037] Skilled artisans will appreciate that elements in the figures are illustrated for simplicity and clarity and have not necessarily been drawn to scale. For example, the dimensions and / or relative positioning of some of the elements in the figures may be exaggerated relative to other elements to help to improve understanding of various embodiments of the present invention. Also, common but well-understood elements that are useful or necessary in a commercial feasible embodiment are often not depicted in order to facilitate a less obstructed view of these various embodiments. It will further be appreciated that certain actions and / or steps may be described or depicted in a particular order of occurrence while those skilled in the art will understand that such specificity with respect to sequence is not actually required. It will further be appreciated that certain actions and / or steps may be described or depicted in a particular order of occurrence while those skilled in the art will understand that such specificity with respect to sequence is not actually required. It will also be understood that the terms and expressions used herein have the ordinary technical meaning as is accorded to such terms and expressions by persons skilled in the technical field as set forth above except where different specific meanings have otherwise been set forth herein.Electronically Filed May 6, 2026DETAILED DESCRIPTION

[0038] Apparatuses, systems, and methods of the present disclosure may, for example, inspect a vial seal based on one or more silhouette images of a vial after completion of an associated vial seal crimping process. For example, such systems and methods may identify a plurality of zones within the silhouette image and analyze the silhouette image to measure a plurality of evaluation metrics for a seal crimp of the vial. The measurements are based on image data within the identified plurality of zones. Such systems and methods may determine whether the seal crimp of the vial includes a defect based on whether at least one of the plurality of evaluation metrics exhibits a corresponding alert condition and reject or accept the vial by determining whether the seal crimp of the vial includes the defect.

[0039] Two-dimensional (2D) vision-based vial seal crimp inspection systems suffer from several significant limitations that result in unacceptable rates of false rejects and false accepts. These systems can be prone to errors arising from illumination conditions and lensing artifacts that prevent clearly delineated boundaries between the seal crimp and background portions of captured images. Additionally, variations in crimp material properties, such as differences in reflectivity across batches of anodized aluminum seals, can cause inconsistent defect detection outcomes when systems configured for one batch are applied to subsequent batches with slight variations in surface finishing or color. The mechanical nature of vial seal crimping processes, combined with ongoing variations in crimping blades and slight dimensional variations in vial lips, can produce seal crimp defects that are difficult to reliably detect. Additionally, the throughput requirements for crimp inspection systems limit the complexity of image processing and analysis.

[0040] The present disclosure addresses these problems by utilizing silhouette imaging techniques that generate high-contrast images with well-defined edges, thereby reducing sensitivity to seal surface reflectivity and color variations. By telecentrically arranging cameras and backlights, the disclosed systems can produce silhouette images that clearly delineate the boundary between the vial seal crimp and the background, enabling and efficient accurate measurement of evaluation metrics such as angles, areas, and distances within defined zones of the silhouette image. Use of zones defined based on a reference pattern can also reduce the amount of image processing so that high throughput inspection can be maintained. Furthermore, the use of multiple cameras oriented at angles relative to each other can provide improved coverage of the seal crimp area without requiring vial rotation during inspection, thereby increasing throughput while maintaining comprehensive defect detection capability and reducing mechanical complexity. The disclosed multi-metricElectronically Filed May 6, 2026 alert conditions can enable fine distinction between defective and non-defective seal crimps, particularly in ambiguous boundary cases where individual metrics alone may be insufficient. Additionally, because the disclosed multi-metric alert conditions can include simple Boolean expressions, they can efficiently determine whether a seal crimp includes a defect while maintaining the high throughput used for crimp inspection.

[0041] Disclosed herein are apparatuses, systems, and methods for vial seal inspection that reduce false rejects and / or false accepts for two-dimensional vision-based vial seal inspection. Apparatuses, systems, and methods disclosedherein can provide for vial seal inspection using clear delineation between background image portions and foreground crimp seal portions. Apparatuses, systems, and methods disclosed herein can provide for for vial seal inspection using multiple camera heads to avoid the vial rotation.

[0042] Turning to FIG. 1, a sealed vial 100a may include a vial 105 (e.g., a glass body, etc.) having a central axis 104, a vial plug 110 (e.g., a rubber seal, etc.), a vial seal 115 (e.g., an aluminum seal, a metal seal, etc.) having an upper surface 118. The vial plug 110 may be at least partially integral with the vial seal 115. A sealed vial may also include a vial flip cap 125. A vial flip cap 125 may be at least partially integral with a vial seal 115.

[0043] A vial 105 may include a vial neck portion 106 having a vial lip 107 with a vial lip lower surface 108 and a vial lip upper surface 109. The sealed vial 100 may include a vial seal crimp surface 119 or skirt that at least partially extends parallel to the lower surface 108 of the vial lip 107 subsequent to the vial seal 115 being crimped onto the vial lip 107.

[0044] With reference to FIGs. 2A-2C, components 200a-c for sealing a vial 205a may include a vial plug 210a, c, a vial seal 215a-c. The vial 205a may include a vial neck portion 206a, a vial lip 207a having a vial lip lower surface 208a and a vial lip upper surface 209a. The vial plug 210a,c may include a vial plug insert portion 21 la, a vial plug lip 212a, a vial plug upper surface 213a, 220c, and a vial plug needle insert area 214a, 220c. The vial seal 215a-c may include a vial seal periphery surface 216a-c, a vial seal lower edge 217a-c, a vial seal upper surface 218b-c, a vial seal inner surface 221c, and a vial seal aperture 222c. After crimping, the vial seal periphery surface 216a-c forms a skirt that includes the lower seal lower edge 217a-c. The vial seal 215a-c may also include an integral vial flip top 225b. The vial flip top 225b may include a central portion connected via a hinge 226b. The central portion of the vial flip top 225b may be flipped to orientation 227b.

[0045] Turning to Fig 3 A, a vial seal crimping system 300a may include at least one high speed rotating blade 338. For information purposes, only some details of the vial seal crimping system are included in FIG. 3A. The vial seal crimping system 300a may be aElectronically Filed May 6, 2026 mechanical process. The metal blade 338 may press on the vial seal 315 of a vial 305 to generate a crimp 324 and a bend 323 on a portion of the vial seal 315 to form a vial seal crimp surface 119 around the vial lip lower surface 108, 208a. Because a vial seal crimping system 300a may include a highly mechanized process, combined with ongoing variations in the blades 338, and slight variations in vial lip 107, 207a dimensions, the vial seal crimping system 300a may be prone to defects with few available process adjustments to the mechanical crimping process to reduce the occurrence of these defects.

[0046] In particular, FIG. 3B shows chart 350 documenting a range of aluminum crimp seals of decreasing quality. The top row of FIG. 3B shows photographs of aluminum crimp seals, while the bottom row of FIG. 3B shows a corresponding x-ray images for each aluminum crimp seal. Quality of crimps can be indicated, for example, by the angle of the crimp seal, which indicates the amount of crimp present. Acceptable crimp types 352 may be classified as crimps 1 through crimp 5 (e.g. C1-C5). Crimp 6 (e.g., C6) may include an intermediate crimp type 354 that will act as a warning to adjust the capping machine. For example, as shown in the x-ray image of C6, C6 has a small angle toward the surface of the glass. Crimps 7, 8 and 9 (e.g., C7-C9) may include unacceptable defective crimp types 356. For example, as shown in the x-ray image of C7, C7 has no angle.

[0047] The apparatuses, systems, and methods of the present disclosure may implement vial seal inspection of the seals applied via the vial seal crimping system 300a as part of a subsequent stage of an assembly line system as described below.

[0048] Figs. 4A and 4B show an example a defect detection station 400 for detecting seal crimp defects in vials (e.g., vial 105, 205a, etc.). The defect detection station 400 may be a part of an assembly line system with at least the vial seal crimping system 300a of FIG. 3. As shown in FIG. 4A, the defect detection station 400 includes a vial transport system 402 and a defect detection system 404.

[0049] The defect detection system 404 shown in FIGs. 4A-4B includes a first camera 406A, a second camera 406B, a first backlight 408A telecentrically arranged with the first camera 406A, a second backlight 408B telecentrically arranged with the second camera 406B, and a control unit 410 that is operatively coupled to receive and process images captured by the first camera 406 A and the second camera 406B. Each of the first camera 406A and the second camera 406B may include respective telecentrically arranged lenses. By telecentrically arranging the cameras 406 and the respective backlights 408, the cameras produce high-contrast image data having well-defined edges. For a telecentrically arranged camera and lenses, rays of light exiting the lens are substantially parallel. For example, aElectronically Filed May 6, 2026 camera and / or lenses can be telecentrically arranged by being arranged at particular distances from a target object so that rays of light exiting the lens are substantially parallel. A benefit of a telecentric arrangement is that it permits imaging a vial at a range of positions with respect to the lens without affecting the precision of distance measurements. As a result, postprocessing techniques that analyze the image data are more likely to accurately detect edges and / or other features of the vial seal. Additionally, by capturing high-contrast image data with well-defined edges, image processing techniques can be more efficient and high throughput.

[0050] The vial transport system 402 may include a star wheel or other mechanical transport device known in the art (e.g., a conveyor system). As shown in FIG. 4A, the vial transport system 402 may include a plurality of recessed openings 412 that secure individual vials in the vial transport system 402 for transport to an imaging location 418. As shown in FIG. 4A, the imaging location 418 may include a location between the first camera 406A and the first backlight 408 A and also the second camera 406B and the second backlight 408B.

[0051] The first camera 406A and the second camera 406B may be configured to operate in conjunction with the first backlight 408A and the second backlight 408B to generate silhouette images (e.g., silhouette images 500 and 700 shown in FIGs. 5 A and 7A) of a vial located at the imaging location 418. As described in more detail below, the silhouette images generated by the first camera 406A and the second camera 406B may define a clear hard boundary of the vial relative to the background. For example, the first camera 406 A and the second camera 406B may generate the silhouette images by assigning background portions of the image a first pixel value (e.g., white or grey value) and non-b ackground portions a second different contrasting pixel value (e.g., black). The first camera 406A and the second camera 406B may distinguish between the background and non-b ackground portions using the intensity of the light produced on the respective sensors of the cameras by the first backlight 408 A and the second backlight 408B. In some embodiments, the first camera 406 A, the second camera 406B, the first backlight 408A, the second backlight 408B, and the control unit 410 may include a Keyence TM-X5000 telecentric measurement and optical comparator system. Furthermore, the first camera 406A, the second camera 406B, the first backlight 408A, the second backlight 408B may be mounted together on a single bracket 420.

[0052] Control unit 410 may generally be configured to control / automate the operation of defect detection system 404, and to receive and process images captured / generated by defect detection system 404 (e.g. by the first camera 406A and the second camera 406B). Control unit 410 may be a general -purpose computer that is specifically programmed toElectronically Filed May 6, 2026 perform the operations discussed herein, or a special-purpose computing device.Furthermore, the Control unit 410 may be operatively coupled directly or indirectly (e.g. via one or more intermediary devices, components, etc.) to the first camera 406A and the second camera 406B via wired or wireless means known in the art.

[0053] As seen in FIG. 4 A, control unit 410 includes a processing unit 422 and a memory unit 424. In some embodiments, however, control unit 410 includes two or more computers, microcontrollers, and / or processing units thereof that are either co-located or remote from each other. In these distributed embodiments, the operations described herein relating to processing unit 422 and memory unit 424, or relating to any of the modules implemented when processing unit 422 executes instructions stored in memory unit 424, may be divided among multiple processing units and / or multiple memory units.

[0054] Processing unit 422 includes one or more processors, each of which may be a programmable microprocessor that executes software instructions stored in memory unit 424 to execute some or all of the functions of control unit 410 as described herein. Processing unit 422 may include one or more graphics processing units (GPUs) and / or one or more central processing units (CPUs), for example. Alternatively, or in addition, one or more processors in processing unit 422 may be other types of processors (e.g., application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), etc.), and some of the functionality of control unit 410 as described herein may instead be implemented in hardware.

[0055] Memory unit 424 may include one or more volatile and / or non-volatile memories. Any suitable memory type or types may be included in memory unit 424, such as read-only memory (ROM) and / or random access memory (RAM), flash memory, a solid-state drive (SSD), a hard disk drive (HDD), and so on. Collectively, memory unit 424 may store one or more software applications, the data received / used by those applications, and the data output / generated by those applications. For example, memory unit 424 stores the software instructions of various modules that, when executed by processing unit 422, perform various functions for the purpose of processing the images by the first camera 406A and the second camera 406B to identify defective seal crimps as described in more detail below.

[0056] As noted above, control unit 410 may be a distributed system, in which case one, some, or all of the functions of the control unit 410 may be implemented in whole or in part by a different computing device or system (e.g., by a remote server coupled to control unit 410 via one or more wired and / or wireless communication networks). Moreover, the functionality of any modules of the control unit 410 may be divided among different softwareElectronically Filed May 6, 2026 applications. Furthermore, as shown in FIG. 4A the control unit 410 may interface by wired or wireless means with an assembly line controller 426. The assembly line controller 426 may include a processing unit and memory unit similar to the processing unit 422 and memory unit 424 described above. The assembly line controller 426 may be configured to operate the vial transport system 402 and other stations and components of the assembly line system. For example, the assembly line controller 426 may reject or accept a vial being imaged by the defect detection system 404 based on signals (e.g., an alert signal, pass signal, etc.) generated by the control unit 410.

[0057] With reference to FIG. 4C, a coverage graph 428 of a first field of view 430A for the first camera 406 A and a second field of view 430B for the second camera 406B with respect to the imaging location 418 is shown. The two cameras 406A, 406B can be arranged to increase the coverage of the imaging location and reduce coverage gaps. In particular, the coverage graph 428 shows the first field of view 430A and the second field of view 430B when the first camera 406A and the second camera 406B are oriented relative to each other at an angle of approximately 90 degrees. As shown in FIG. 4C, for two cameras oriented at an angle of approximately 90 degrees, the first field of view 430A may cover 86 degrees of the imaging location 418 from the front view of the lens of the first camera 406 A and a complimentary 86 degrees from the back of the vial relative to the lens of the first camera 406 A by virtue of the shadow area of the vial. The second field of view 430B may cover a non-overlapping different 86 degrees of the imaging location 418 relative to front and back view location of the vial for the second camera 406B. Furthermore, the coverage graph 428 for the 90-degree arrangement of the cameras results in a coverage gap region 432 of approximately 4 degrees.

[0058] It should be appreciated that alternative respective arrangements of the first camera 406A and the second camera 406B will produce different coverage of the imaging location 418. For example, FIG. 4D shows an arrangement of the first camera 406 A and the second camera 406B for use in another vial transport 434 having a smaller diameter than the vial transport system 402 of FIG. 4A. In this embodiment, the first camera 406A and the second camera 406B are arranged at approximately 70 degrees to each other on an alternative bracket 436. FIG. 4E shows another coverage graph 438 of a first field of view 440A for the first camera 406A and a second field of view 440B for the second camera 406B with respect to the imaging location 418 for the camera arrangement shown in FIG. 4D. For example, the first field of view 440 A may cover 86 degrees of the imaging location 418 from the front and rear views of the vial by the lens of the first camera 406A and the second field of view 440BElectronically Filed May 6, 2026 may cover a different front and rear 86 degrees of the imaging location 418 with an overlapping area 442 of approximately 15.98 degrees. Furthermore, the coverage graph 438 for the 70-degree arrangement of the cameras results in a coverage gap region 444 of approximately 23.98 degrees.

[0059] Regardless of the specific arrangement of the first camera 406A and the second camera 406B, use of two cameras can provide an increased coverage of the crimping area that may not be possible with a single camera in cases where the vial is not rotated at the time of inspection. Increased coverage is beneficial because a seal crimp defect may occur on portions of the vial seal hidden from view of a single camera. By using two or more stationary cameras oriented at an angle with one another, increased coverage can be achieved without the need for complicated mechanical movement such as rotation. In some embodiments, the first camera and second camera are oriented relative to each other at an angle in a range of 65-95 degrees. In some embodiments, the first camera and second camera are oriented relative to each other at an angle of about 70 degrees. In some embodiments, the first camera and second camera are oriented relative to each other at an angle of about 90 degrees.

[0060] Operation of the defect detection system 404 to detect seal crimp defects in a vial located at the imaging location 418 will be described in more detail with reference to FIGs.5A-5M and 6A-6Q. For clarity, reference will also be made to the FIGs. 4A-4E and the components of the defect detection system 404 throughout this section.

[0061] FIG. 5A shows an example silhouette image 500 depicting a silhouette 502 for a vial with a defect free seal crimp. In operation, the first camera 406A or the second camera 406B generate the silhouette image 500 and send the silhouette image 500 to the control unit 410 for processing. It should be appreciated that in some embodiments silhouette image 500 may instead be generated by the control unit 410 using known illumination intensity output by the first backlight 408A or second backlight 408B and raw readouts of image sensor pixel values from the first camera 406 A or the second camera 406B.

[0062] In either case, after the control unit 410 receives the silhouette image 500, the processing unit 422 analyzes the silhouette image 500 to determine whether a seal crimp defect is present or not. To perform this analysis, the processing unit 422 can first identify a plurality of zones within the silhouette image 500. The plurality of zones can be defined based on a reference pattern associated with a crimp type for the vial being imaged and evaluated for possible seal crimp defects. The processing unit 422 may retrieve the reference pattern from the memory unit 424 or another suitable data storage system operatively coupledElectronically Filed May 6, 2026 to the processing unit 422.

[0063] FIG. 5B shows an example reference pattern 504 that includes 6 zones 506A, 506B, 508A, 508B, 510A, and 510B. In general, the different zones of the reference pattern 504 may be referenced to different portions of a seal crimp within silhouette images generated using the defect detection system 404. Referencing the different zones of the reference pattern 504 to the different portions of the seal crimp within the silhouette image may include initially defining size and the relative location of each of the different zones within a reference silhouette image to identify the portion of the seal crimp to be referenced within that zone. In this way, the different zones can be defined to indicate the expected location of the referenced portion the seal crimp within silhouette images of vials generated by the defect detection system 404. During operation, the control unit 410 may attempt to locate the zones within the silhouette image 500 such that they match the relative locations and sizes as defined within a predefined tolerance. In some embodiments, the control unit 410 may attempt to position the zones of the reference pattern 504 within a generated silhouette image and determine that the generated silhouette image does not match the reference pattern 504 when at least one of the zones cannot be aligned with the defined relative location and / or size for that zone.

[0064] In the example shown in FIG. 5B, the zones 506A and 506B are top zones referenced to a top portion of a flip-off portion of the seal crimp on a left and right side of the image, respectively. The zones 508A and 508B are side zones referenced to a side portion of the seal crimp on the left and right sides of the image, respectively. Finally, the zones 510A and 510B are bottom zones referenced to a bottom portion of the left and right sides of the image, respectively, where the crimp is tucked under the lip of the vial.

[0065] When the processing unit 422 identifies the plurality of zones of the reference pattern 504 in the silhouette image 500, the processing unit 422 may proceed to analyze the silhouette image 500 to generate a plurality of evaluation metrics for the seal crimp of the vial. The generation of the evaluation metrics may be based on measurements of vial seal features represented by the image data and within the identified plurality of zones. For example, the measurements may be measured with respect to reference locations that the processing unit 422 identifies within the zones of the reference pattern 504.

[0066] As described in more detail below, the evaluation metrics and the reference locations may be defined and associated with the reference pattern 504 in the memory unit 424 or alternative data storage system. To define the reference pattern associated with the crimp, the processing unit 422 may generate a representative silhouette image of a crimpElectronically Filed May 6, 2026 defect-free vial having the crimp type and defining a size and relative location of the plurality of zones within the representative silhouette image. The defined zones may then be saved in the memory unit 424 or other data storage system as the reference pattern associated with the crimp type. The representative silhouette image may be generated based on image data generated by the first camera 406A or 406B, similar camera systems of a testing system distinct from the assembly line system described herein, or from existing image libraries of defect free seal crimps. Once the reference pattern 504 is defined, the plurality of evaluation metrics may be defined (e.g., by user input via a vial inspection configuration application or automatically by the inspection system) with respect to the plurality of zones and associate with the reference pattern.

[0067] Example reference locations are shown in FIGs. 5C-5M. Each image in FIGs. 5A-5B and 5G-5L shows the same silhouette of a vial, with each image showing a different reference locations on that silhouette, while each image in FIGs. 5C-5F and 5M shows the same silhouette of a vial, with each image showing a different reference locations on that silhouette.

[0068] FIG. 5C shows a location of a crimp region 511 A and a flip top region 51 IB. FIG. 5D shows a top edge location 512C of the silhouette 502. FIGs. 5E and 5F show top edge locations 512A and 512B identified within the top zones 506A and 506B of FIG. 5B, respectively. FIGs. 5G and 5H show side edge locations 514A and 514B identified within the side zones 508 A and 508B of FIG. 5B, respectively. FIGs. 51 and 5 J show bottom edge curvature locations 516A and 516B identified within the bottom zones 510A and 510B of FIG. 5B, respectively. FIGs. 5K and 5L show maximum bottom edge locations 518A and 518B identified within the bottom zones 510A and 510B of FIG. 5B, respectively. The maximum bottom edge locations 518A and 518B may comprise the lowest edge portion of the bottom zones 510A and 510B that is attributable to the seal crimp of the vial and are different from the curvature locations 516A and 516B, which may be aligned with the left and rightmost edges of the seal crimp of the vial present within the bottom zones 510A and 510B. For example, the bottom edge can be identified at different locations along the crimp skirt. FIG. 5M shows intersection 520A, 520B of the side edges 514A, 514B and bottom edges 518A, 518B. It should be appreciated that the depicted zones and locations represent one example reference pattern and other reference patterns may be implemented in other systems and / or tailored to other vial types.

[0069] The processing unit 422 may use various methods or tools (e.g., locator tools) to identify the edge locations 512A, 512B, 514A, 514B, 516A, 516B, 518A, and 518B. InElectronically Filed May 6, 2026 general, the processing unit 422 scans each zone for changes in pixel values that identify the corresponding edge location(s) associated with each zone. For example, the processing unit 422 may scan for polarity changes in the pixel values, use sobel or gradient edge detection methods, use intensity thresholding (e.g., blob analysis), use straight line fit (e.g. ‘rake’ algorithm) methods, etc. to identify the edge location. In some embodiments, the processing unit 422 may identify the edge location by connecting the results of individual point edge detections.

[0070] An example of a pattern tool is described with respect to FIG. 5C. A pattern tool can be used to identify a crimp region 511 A and a flip top region 51 IB. This pattern tool can contribute to consistent positioning and a stable region of interest for subsequent measurements. For example, a pattern tool can provide an anchor, such as the center of the pattern, which can be used to measure angles and distances. A pattern tool can receive one or more reference images non-defective vials to create a pattern area. A pattern tool with a search region and the pattern area can search for a pattern in the search region. For example, a pattern tool can perform a correlation of the one or more reference images with a sample image.

[0071] An example of a top edge locator is described with respect to FIGs 5D-5F. As shown in FIG. 5D, a rectangular region of interest can be placed across the top of the flip off or cap, and an edge tool within the region of interest can detect the top surface 512C, thereby establishing a stable vertical reference for height or distance measurements. Alternatively or additionally, as shown in FIGs. 5E-5F, a line tool can locate the top edge 512A, 512B on the left and right sides, respectively, by scanning vertically for polarity changes.

[0072] An example of a side edge locator is shown with respect to FIGs. 5G-5H. As shown in FIGs. 5G-5H, the left side edge 514A and right side edge 514B can be found using a line tool scanning horizontally for polarity changes. Additionally or alternatively, a vertical region of interest can be used to find each side edge.

[0073] An example of a bottom edge locator is shown with respect to FIGs. 5I-5L. As shown in FIGs. 5I-5L, the bottom edges 516A, 516B, 518A, 518B can be detected using a line tool scanning vertically for polarity changes. A bottom edge can be detected at different locations along the skirt.

[0074] An example of a locator of the intersection of the side edge and bottom edge is shown with respect to FIG 5M. FIG. 5M shows the intersection 520A of the left side edge 514A and the left bottom edge 518A and the intersection 520B of the right side edge 514B and the right bottom edge 518B. These intersections are marked with an X at a transitionElectronically Filed May 6, 2026 between the vial body and the crimp skirt. These intersections 520A, 520B can be used as anchors for distance measurements from the skirt and / or as geometric references for arc and radius tools and for some advanced calculations.

[0075] After the processing unit 422 identifies the edge locations 512A, 512B, 514A, 514B, 516A, 516B, 518A, and 518B, the processing unit 422 may make measurements utilized by the plurality of evaluation metrics for use in determining whether the seal crimp of the vial being imaged is or is not defective. Example evaluation metrics will be described in connection with FIGs. 6A-6Q. Each image in FIGs. 6A-6D and 6I-6J shows the same silhouette of a vial, with each image showing a different evaluation metric on that silhouette, while each image in FIGs. 6E-6H and 6K-6Q shows the same silhouette of a vial, with each image showing a different evaluation metric on that silhouette.

[0076] For example, FIGs 6 A and 6B show area evaluation metrics 600 A and 600B that the processing unit 422 determines for the bottom zones 510A and 510B of FIG. 5C, respectively. FIG. 6A shows evaluation metric 600A on the left side of the silhouette of the vial, and FIG. 6B shows evaluation metric 600B on the right side of the silhouette of the vial. The area evaluation metrics 600A and 600B correspond to the area of the bottom zones 510A and 510B occupied by pixels associated with the seal crimp of the vial. In some embodiments, the processing unit 422 may measure the area evaluation metrics 600A and 600B by counting the number of black pixels and converting the count into an area value using a known size of the individual pixels. In the example in FIGs. 6A-6B, the area evaluation metric is 0.74 mm2on the left side, and 0.66 mm2on the right side.

[0077] FIGs. 6C and 6D show an example of angle evaluation metrics 602 A and 602B. Angle evaluation metrics can quantify how far and how uniformly the aluminum skirt is folded around the vial lip. As a result, angle metrics can indicate loose crimps, missing crimps, and / or asymmetric crimps. FIG. 6C shows evaluation metric 602A on the left side of the silhouette of the vial, and FIG. 6D shows evaluation metric 602B on the right side of the silhouette of the vial. The angle metric 602 A may be measured between the side edge location 514A and bottom edge location 516A and the angle evaluation metric 602B may be measured between the side edge location 514B and the bottom edge location 516B. In the example in FIGs. 6C and 6D, the angle evaluation metric is 107.5° on the left side and 112.0° on the right side.

[0078] In the example in FIGs. 6C and 6D, the angle evaluation is a global angle evaluation metric that indicates the overall wrap of the crimp skirt. In this example, on the left side, a large arc is defined between side edge location 514A (e.g., a vertical reference atElectronically Filed May 6, 2026 the vial sidewall) and bottom edge location 516A (e.g., a line that follows the crimped skirt down to the neck or lip). On the right side, a large arc is defined between side edge location 514B and bottom edge location 516B. This global angle evaluation metric can return an overall angle through which the skirt wraps around the lip (e.g., -110-140° in the example). A global evaluation metric can indicate crimp defects, including loose or incomplete crimps (e.g., crimps where the skirt stays more vertical, so the side-to-crimp angle becomes smaller), over-crimps (e.g., crimps where the skirt is pulled further under the lip, so the angle becomes larger than the target range), and asymmetry (e.g., differences between left and right angles indicating that one side is tighter or looser than the other)

[0079] FIGs. 6E-6H show an example where multiple angle evaluation metrics are measured at different heights along the skirt (e.g., local angle evaluation metrics). By measuring angles at multiple heights, a more detailed shape of the crimp can be evaluated, and more ambiguous defects can be identified. In this example, the angle is measured at four different heights along the skirt on each side of the vial. On the left side, an angle 602A1, 602A2, 602A3, 602A4 is measured between a side edge location 514A and a line tangent to the skirt edge at each height 516A1, 516A2, 516A3, 516A4. On the right side, an angle 602B1, 602B2, 602B3, 602B4 is measured between a side edge location 514B, and a line tangent to the skirt edge at each height 516B1, 516B2, 516B3, 516B4. Together, multiple local angle evaluation metrics can describe a shape or profile of the crimp skirt as it bends under the lip of the vial. Generally, for a non-defective crimp, the local angle evaluation metrics follow a characteristic progression from smaller angles near the side edge to larger angles near the fold. Angles from Figures FIGs. 6E-6H, can be combined, for example, using standard deviation calculation. A standard deviation of multiple angle evaluation metrics can indicate a missing crimp, with a low standard deviation indicating a missing crimp. Using multiple angle evaluation metrics can indicate crimp defects, including a missing crimp (e.g., a low standard deviation among multiple angle measurements), over-crimping or a sharp fold (e.g., the skirt is pulled strongly under the lip, the local angles becoming larger, with the change in angle from one height to the next more abrupt), shape anomalies (e.g., comparing the pattern of angles (Angle 1 — Angle 4) on the left vs right side to detect local deformities or skewed crimps even if the global angle evaluation metric is acceptable).

[0080] FIG. 6E shows a first local angle 602A1, 602B1 at the lower skirt region. This local angle is an angle between the side edge 514A, 514B and a tangent 516A1, 516B1 at the beginning of the skirt bend under the lip of the vial. In this example, the local angle evaluation metric is 28.0° on the left side (602A1) and 28.8° on the right side (602B1). TheElectronically Filed May 6, 2026 angle at this position can be sensitive to early pull-down of the crimp. For example, small angles (e.g., smaller than typical) at this location can indicate a loose or missing crimp

[0081] FIG. 6F shows a second local angle 602A2, 602B2 at the mid skirt region. This local angle is an angle between the side edge 514A, 514B and a tangent 516A2, 516B2 at the mid-skirt region. In this example, the local angle evaluation metric is 54.8° on the left side (602A2) and 55.6° on the right side (602B2). For a non-defective crimp, angles at this position generally characterize the normal progression of the skirt as it folds under the lip. Deviations from the typical range of angles, especially when combined with deviations at other locations, can indicate loose or over-crimped seals.

[0082] FIG. 6G shows a third local angle 602 A3, 602B3 at near the upper skirt region, close to the fold. This local angle is an angle between the side edge 514A, 514B and a tangent 516A3, 516B3 at the upper-skirt region. In this example, the local angle evaluation metric is 70.9° on the left side (602A3) and 72.6° on the right side (602B3). Larger angles at this location relative to the first and second location can confirm that the skirt has been pulled sufficiently under the lip, whereas reduced values can indicate an under-folded or loose crimp. Additionally, deviations from the typical range of angles at this location, especially when combined with deviations at other locations, can indicate loose or over-crimped seals.

[0083] FIG. 6H shows a fourth local angle 602A4, 602B4 at the most folded portion of the skirt. This local angle is an angle between the side edge 514A, 514B and a tangent 516A4, 516B4 at the most folded portion. In this example, the local angle evaluation metric is 73.7° on the left side (602A4) and 76.9° on the right side (602B4). This angle can indicate how tightly the skirt is tucked under the lip. For example, angle measurements at this location can be especially useful for distinguishing borderline loose crimps from fully acceptable crimps. Additionally, this measurement can feed into advanced angle-sum calculations.

[0084] FIGs. 61 and 6J show an example of distance evaluation metrics 604A and 604B. FIG. 61 shows evaluation metric 604A on the left side of the silhouette of the vial, and FIG.6J shows evaluation metric 604B on the right side of the silhouette of the vial. The distance evaluation metric 604A may be measured between the top edge location 512A and bottom edge location 518A and the distance evaluation metric 602B may be measured between the top edge location 512B and the bottom edge location 518B. For the example in FIGs. 61 and 6J, the distance evaluation metric is 8.791 mm for the left side (604A) and 8.870 mm for the right side (604B). Distance evaluation metrics 604A, 604B can indicate a distance from the flip-off to crimp, which can indicate whether there is a missing crimp. For example, a distance below a threshold can indicate a missing crimp.Electronically Filed May 6, 2026

[0085] FIGs. 6K and 6L show an example of arc radius evaluation metrics 605A and 605B that indicate a radius of an arc 610A, 61 OB at the curved tuck-under region of the crimp. FIGs. 6K and 6L show an image of the same vial taken by two different cameras oriented relative to each other at an angle of 60 degrees. An arc can be fitted to this region by defining two edge locations (e.g., the side edge 514A, 514B and bottom edge 516A, 516B), and the radius of curvature of this arc can be measured. For the example in FIG. 6K, the arc radius evaluation metric is 0.758 mm for the left side (605A) and 0.792 mm for the right side (605B). For the example in FIG. 6L, the arc radius evaluation metric is 0.841 mm for the left side (605A) and 0.842 mm for the right side (605B). The radius of curvature can indicate how loose or tight the crimp is. For example, a larger radius can indicate a more open skirt (e.g., looser crimp), whereas a smaller radius can indicate a tighter fold (e.g., stronger crimp). Radius evaluation metrics can be used directly or in advanced calculations (e.g., in multimetric alerts) to help distinguish marginal loose crimps from acceptable crimps.

[0086] FIGs. 6M-6Q show an example of intersection distance evaluation metrics 606A1, 606B1, 606A2, 606B2, 606A3, 606B3. This intersection evaluation metric can be measured at different locations, as shown in FIG. 6M. The rectangles in FIG. 6M show different sampling positions 612A1, 612A2, 612A3, 612A4, 612B5, 612B2, 612B3, 612B4, 612B4, 612B5 along the skirt curve where the intersection point can be measured. FIGs. 6N-6Q show intersection distance evaluation metrics for four of the five sampling positions shown in FIG. 6M.

[0087] As shown in FIGs. 6N-6Q, an intersection distance evaluation metric can be determined by finding an intersection a intersection 520A, 520B between a side edge 514A, 514B, and a bottom edge (e.g., at a vial-to-skirt transition) 518A, 518A at location 612A5, 612B5 and measuring a linear distance along a defined direction to the crimp edge at different sampling positions. In this example, the same intersection point 520A, 520B is used for each intersection distance evaluation metric. The crimp edge 516A1, 516A2, 516A3, 516A4, 516B1, 516B2, 516B3, 516B4 can be defined at the different sampling positions 612A1, 612A2, 612A3, 612A4, 612B1, 612B2, 612B3, 612B4, 612B4. For each sampling position 612A1, 612A2, 612A3, 612A4, 612B1, 612B2, 612B3, 612B4, 612B4, the distance is measured between the intersection 520A, 520B and the respective crimp edge 516A1, 516A2, 516A3, 516A4, 516B1, 516B2, 516B3, 516B4 at that location. This intersection distance evaluation metric can indicate or quantify how far the skirt of the crimp is pulled in on each side. For example, an increase in distance or an increase in asymmetry between the left and right side can indicate loose crimps and / or skewed or uneven crimping. FIG. 6NElectronically Filed May 6, 2026 shows a distance between intersection 520A, 520B and crimp edge 516A1, 516B1 at a first sampling position 612A1, 612B1; FIG. 60 shows a distance between intersection 520A, 520B and crimp edge 516A2, 516B2 at a second sampling position 612A2, 612B2; FIG. 6P shows a distance between intersection 520A, 520B and crimp edge 516A3, 516B3 at a third sampling position 612A3, 612B3; FIG. 6Q shows a distance between intersection 520A, 520B and crimp edge 516A4, 516B4 at a fourth sampling position 612A4, 612B4. Table 1 summarizes the intersection distance evaluation metrics as determined in FIGs. 6N-6Q. Table 1. Intersection Distance Evaluation Metrics0.309 mm (606A1), 0.330 mm (606B1)0.324 mm (606A2), 0.338 mm (606B2)3 (612A3, 612B3)4 (612A4, 612B4)

[0088] It should be appreciated that more or fewer evaluation metrics may be used. Furthermore, the plurality of evaluation metrics may be dependent on the type of crimp seal being examined. For example, FIG. 7A shows another silhouette image 700 for a silhouette 702 of an alternative vial and crimp seal arrangement the silhouette image 500 of FIG. 5 A. For example, the alternate crimp seal arrangement may include a 20R vial as opposed to the silhouette image 500 depicting a 50R vial. FIG. 7B shows evaluation metrics that the processing unit 422 may assess for the 20R vial of the silhouette images 700. For example, the processing unit 422 may identify side edges 704A and 704B and bottom edge curvature locations 706A and 706B in a process similar to that described above for side edges 514A and 514B and bottom edge curvature locations 516A and 516B. The evaluation metrics for the 20R vial may include angle 708A measured between the side edge 704A and bottom curvature location 706A and angle 708B measured between the side edge 704B and bottom curvature location 706B. In some embodiments, the evaluation metrics for the 20R vial may also include a difference between the angle 708A and the angle 708B.

[0089] After the processing unit 422 generates the evaluation metrics for the vial and seal crimp being imaged, the processing unit 422 determine whether any of the plurality of evaluation metrics exhibit a corresponding alert condition and reject or accept the vial byElectronically Filed May 6, 2026 determining whether the seal crimp of the vial includes a defect based on the evaluation metrics and the alert condition. For example, some or all of the evaluation metrics may have a corresponding absolute alert condition that represents a primary threshold value for which the processing unit 422 will always indicate that the seal crimp is defective. In the case of the 50R vial shown in the silhouette image 500 of FIG. 5 A, the area evaluation metrics 600A and 600B may have an primary threshold value of greater than 1.72, 1.70, or 1.74 square millimeters and the angle evaluation metrics 602A and 602B may have a primary threshold of greater than 165, 162, or 167 degrees. In the case of the 20R vial shown in the silhouette image 700 of FIG. 7A, the angle evaluation metrics 708 A and 708B may have a primary threshold of greater than 142, 140, or 144 degrees. These thresholds are merely examples can may differ based on the type of vial and the crimp equipment. Appropriate thresholds can be determined based on typical evaluation metrics for defective and non-defective crimps. Depending on the evaluation metric, a value above the primary threshold or below the primary threshold (e.g., outside the threshold) indicates that the seal crimp includes a defect. For example, an alert condition can include values outside a range set by the primary threshold, values above a primary threshold, or below a primary threshold. An alert condition based on an individual evaluation metric (an absolute alert condition) can identify defective crimps that are unambiguous.

[0090] Furthermore, in some embodiments, the processing unit 422 may evaluate whether the plurality of evaluation metrics satisfy a multi-metric alert condition. Utilizing a multi-metric alert condition in addition to the absolute alert condition described above can enable the defect detection system 404 to finely distinguish between defective and nondefective seal crimps. For example, when a boundary between non-defective and defective crimps is more ambiguous based on individual evaluation metrics, a multi-metric alert condition can be used to distinguish between defective and non-defective crimps. For example, the use of the multi-metric alert condition enables the defect detection system 404 to identify defective vials as a function of multiple evaluation metrics rather than relying on a single metric. For example, a seal crimp may be defective when multiple angle, area, or distance values are outside a secondary threshold value that is different from the primary threshold value for the same metric but a seal crimp with similar angle, area, or distance values at only one or two locations may not be defective. Depending on the evaluation metric, values above the secondary threshold or below the secondary threshold (e.g., outside the threshold) indicates that the seal crimp includes a defect. For example, an alert conditionElectronically Filed May 6, 2026 can include values outside a range set by the secondary threshold, values above a secondary threshold, or below a secondary threshold

[0091] A multi-metric alert condition may define a combination of values for the evaluation metrics that are different from and / or below those of the absolute alert condition value but that as a combination indicate that the seal crimp is defective. For example, in the case of the 50R vial shown in the silhouette image 500 of FIG. 5 A, the multi -metric alert condition may depend on a review of the evaluation metrics across the silhouette images captured by both the first camera 406A and the second camera 406B. For example, the multimetric alert condition may include cases where the sum of all four of the distance evaluation metrics (e.g., distance metrics 604A and 604B as measured on both sides of both silhouette images) is greater than 36.6 mm and at least 4 of the area and angle evaluation metrics (e.g., area evaluation metrics 600A, 600B and angle evaluation metrics 602A, 602B as measured in both silhouette images) exceed or be outside corresponding secondary alert thresholds. The secondary alert threshold for the angle metrics may include values greater than 150 degrees and the secondary alert threshold for the area metrics may include values greater than 1.5 square millimeters. In the case of the 20R vial shown in the silhouette image 700 of FIG. 7A, the multi-metric alert condition may be a function of the angle and angle difference metrics in the silhouette images captured by both the first camera 406 A and the second camera 406B. For example, the multi-metric alert condition may include cases where any angle metric is greater than 132 degrees and the difference between the angle 708 A and angle 708B in either image is greater than 16 degrees. These examples of multi -metric alert conditions are not limiting.

[0092] In some embodiments, the processing unit 422 may average or otherwise combine the corresponding evaluation metrics measured from the silhouette image captured by the first camera 406 A and the second camera 406B to obtain a composite metric. For example, the processing unit 422 may average or otherwise combine the angle evaluation metric 602A from the image from the first camera 406A with the measurement from the second camera 406B. An average value may then be compared to the relevant alert thresholds described above or to a different alert threshold determined for the average value specifically.

[0093] In general, the absolute and multi-metric alert conditions correspond to cases where the conditions of the seal crimp under review are likely to represent a defective crimp (e.g., where the vial seal 315 of FIG. 3 is insufficiently wrapped around the vial lip lower surface 108, 208a to prevent leaks of the vial contents or unintended opening of the vial). For example, large area, angle, and distance values would correspond to cases where the amountElectronically Filed May 6, 2026 of seal crimp material present in the lower half of the silhouette images (e.g., the bottom zones 510A and 51 OB) is too great to be consistent with a non-defective seal crimp (e.g., where the vial seal material will be expected to be closely adhered to the vial lip lower surface). To identify the values and conditions that correspond to such cases, the systems and methods described herein may include an iterative process to identify and define such values and conditions using measurements taken from known defective and non-defective vials and or simulations of such vials. The details of this process are described in more detail below in connection with FIGs. 9A-9B and 10A-10C.

[0094] In some embodiments, the processing unit 422 may also determine that the vial has a defective seal crimp when the silhouette image from the first camera 406A or second camera 406B fails to match the reference pattern (e.g., reference pattern 504) associated with the crimp type. For example, FIGs 8A-8C show examples silhouette images where the processing unit 422 fails to match the reference pattern 504, and this failure may trigger rejection of the vial. For example, FIG. 8A shows a silhouette image 800A for a vial with a missing stopper portion where the processing unit 422 fails to match the reference pattern 504. FIG. 8B shows a silhouette image 800B for a vial with a missing cap portion where the processing unit 422 fails to match the reference pattern 504. FIG. 8C shows a silhouette image 800C for a vial with a missing flip-off portion where the processing unit 422 fails to match the reference pattern 504. In some embodiments, the sensitivity of the pattern matching may be set by user input. For example, a minimum and maximum pattern matching parameter can be adjusted, where higher the values result in stricter is the pattern matching. In some embodiments, the pattern matching parameter used to differentiate the cases of no flip off, no cap or no stopper shown in FIGS. 8A-8C may be 70%.

[0095] With reference now to FIG. 9A, a flow chart for a method 900 used to determine absolute alert threshold values (e.g., primary threshold values) for the evaluation metrics to use for the absolute alert condition. The method 900 may be performed using the defect detection system 404 and the control unit 410, similar systems of a testing system distinct from the assembly line system described herein, and / or using existing image libraries of defective and non-defective seal crimps.

[0096] At block 902, the method 900 includes generating, a representative set of defective values for each evaluation metric using defective test vials and a representative set of non-defective values for each evaluation metric using non-defective test vials. These values may then generally be used to define the alert conditions for the plurality of evaluation metrics as a threshold value that delineates the representative set of defective values and theElectronically Filed May 6, 2026 representative set of non-defective values for each of the evaluation metrics. The representative set of defective values and the representative set of non-defective may be generated by generating a first set of testing silhouette images for defect free vials (e.g., vials with acceptable crimp types 352 shown in FIG. 3B) having the crimp type and generating a second set of testing silhouette images for defective crimp vials (e.g., vials with unacceptable defective crimp types 356 shown in FIG. 3B) having the crimp type. These images may then be processed using the methods described above to generate the measured evaluation metrics, which will constitute the representative set of defective values and the representative set of non-defective. The vials used when generating the first set of silhouette images and the second set of silhouette images may include actual defective and non-defective seal crimp vials or simulated defective and non-defective seal crimp vials (e.g., 3D printed or manufactured vials that will produce silhouette images akin to the defective seal crimp such as C7-C9 shown in FIG. 3B and non-defective seal crimps such C1-C5 as shown in FIG. 3B).

[0097] Table 2 below shows an example representative set of non-defective values and Table 3 shows an example representative set of defective values for three 50R type vials and seal crimp e.g., the vial shown in silhouette image 500). The non-defective vials are taken from the groups C1-C3 of the acceptable crimp types 352 shown in FIG. 3B and the defective vials are taken from the groups C8 and C9 of the unacceptable defective crimp types 356 shown in FIG. 3B. When using the defect detection system 404 the process may be conducted multiple times (e.g., 10 times) to obtain the angles, area, and distances measurements for each vial and the position of each vial may be periodically changed to obtain position variability.Table 2 - representative set of non-defective valuesVIAL 1 VIAL 2 VIAL 3 READINGS READINGS READINGS ANGLE 120, 121,122,123 121,122,123,124 122,123,124,125 AREA 50,51,52,53 50,52,53,54 52,53,54,55 DISTANCE 100,101,102,103 101,102,103,104 102,103,104,105Table 3 - representative set of defective valuesVIAL 1 VIAL 2 VIAL 3 READINGS READINGS READINGS ANGLE 140,141,142,143 141,142,143,144 142,143,144,145 AREA 60,61,62,63 61,62,63,64 62,63,64,65Electronically Filed May 6, 2026 DISTANCE 200,201,202,203 201,202,203,204 202,203,204,205

[0098] At block 904, the method 900 includes selecting the maximum value in the set of defective values for each evaluation metric and defective test vial to generate a defect value candidate set for each evaluation metric. The results of the selection process on the examples values of Table 3 are shown below in Table 4. Specifically, Table 4 shows the maximum reading of each vial for each measurement from Table 3.Table 4 - defect value candidate set for each evaluation metricVIAL 1 MAX VIAL 2 MAX VIAL 3 MAX ANGLE 143 144 145AREA 63 64 65DISTANCE 203 204 205

[0099] At block 906, the method 900 includes selecting the maximum value in the set of non-defective values for each evaluation metric and non-defective test vial to generate a nondefective value candidate set for each evaluation metric. The results of the selection process on the examples values of Table 2 are shown below in Table 5. Specifically, Table 5 shows the maximum reading of each vial for each measurement from Table 2.Table 5 - non-defective value candidate set for each evaluation metricVIAL 1 MAX VIAL 2 MAX VIAL 3 MAX ANGLE 123 124 125AREA 53 54 55DISTANCE 103 104 105

[0100] At block 908, the method 900 includes generating an absolute alert threshold for the evaluation metrics, for example, by determining a midpoint between the minimum value in the defect value candidate set and the maximum value in the non-defective value candidate set. In other embodiments, the alert threshold may be higher or lower than the midpoint based on a desired alert sensitivity. In some embodiments, the absolute alert threshold may only be set for a subset of the evaluation metrics. For example, in the case of the 50R type vial, only the angle and area metrics may have an absolute alert threshold. When the operation of block 908 is applied with respect to the example values shown in Tables 3 and 4, the minimum defect angle value is 143, the minimum defect area value is 63, the maximum non-defectiveElectronically Filed May 6, 2026 angle value is 125, and the maximum non-defective area value is 55, which after averaging results in an absolute angle alert threshold of 134 and an absolute area threshold of 59.

[0101] It should be appreciated that the values described above with respect to Tables 2-5 and blocks 902-908 are example values that illustrate the process of the method 900 with respect to example 50R type vials and the method 900 may be utilized to generate other values for similar or different types of vials, seal crimps, and evaluation metrics (e.g., the 20R vial shown in silhouette image 700 of FIG. 7 A).

[0102] With reference now to FIG. 9B, a flow chart for a method 950 used to determine components of the multi-metric alert condition based on two or more of the evaluation metrics. The method 950 may be performed using the defect detection system 404 and the control unit 410, similar systems of a testing system distinct from the assembly line system described herein, and / or using existing image libraries of defective and non-defective seal crimps. The components of the multi-metric alert condition may include a selected composite alert metric, a composite metric alert threshold, secondary thresholds for the plurality of evaluation metrics, an instance threshold, or a combination thereof. The composite alert metric may be a metric that incorporates values from multiple different images of a single vial (e.g., a distance sum of all distances measured in the silhouette images from both the first camera 406A and the second camera 406B) and the composite metric alert threshold may be a respective alert threshold used with respect to the composite alert metric. The secondary thresholds are additional different threshold values for the plurality of evaluation metrics from the primary thresholds for the same metrics that are used to trigger the absolute alert condition. The secondary thresholds may include the threshold values that the control unit 410, in operation of the defect detection system 404, will compare to each of the plurality of metrics to determine whether an alert instance is present. The instance threshold may include the number of alert instances sufficient to trigger the multi-metric alert condition. As one example, the control unit 410 may determine that the multi -metric alert condition exists when a threshold number (e.g., the instance threshold) of the angle or area metrics satisfy respective angle and area thresholds (e.g., the secondary thresholds) and a sum of the distance measurements (e.g., the composite metric) satisfies a respective distance sum threshold (e.g., the composite metric alert threshold).

[0103] At block 952, the method 900 includes generating the representative set of defective values for each evaluation metric using defective test vials and generating the representative set of non-defective values for each evaluation metric using non defective test vials in the manner described above for block 902 of the method 900. These values may thenElectronically Filed May 6, 2026 generally be used to define the components of the multi-metric alert condition. Table 6 below shows an example representative set of non-defective values and Table 7 shows an example representative set of defective values for three 50R type vials and seal crimp (e.g., the vial shown in silhouette image 500). The non-defective vials are taken from the groups C4 and C5 of the acceptable crimp types 352 shown in FIG. 3B and the defective vials are taken from the group C7 of the unacceptable defective crimp types 356 and vials from the gray zone crimp type 354 shown in FIG. 3B. In this way, the representative sets of defective and non-defective values are different when used to identify the absolute alert conditions and the multi-metric alert conditions. However, it should be appreciated, that in some embodiments, the same sets of values may be used for each process.Table 6 - representative set of non-defective valuesVIAL 1 VIAL 2 VIAL 3 READINGS READINGS READINGS ANGLE 129,130,131,132 130,131,132,133 131,132,133,134 AREA 53,54,55,56 54,55,56,57 55,56,57,58 DISTANCE 100,101,102,103 101,102,103,104 102,103,104,105Table 7 - representative set of defective valuesVIAL 1 VIAL 2 VIAL 3 READINGS READINGS READINGS ANGLE 131,132,133,134 132,133,134,135 133,134,135,136 AREA 56,57,58,59 57,58,59,60 57,58,59,60 DISTANCE 200,201,202,203 201,202,203,204 202,203,204,205

[0104] At block 954, the method 900 includes selecting the composite metric and summing the associated values in the set of defective values and the set of non-defective values such that each test vials has an associated defect composite value and an associated non-defective composite value. For purposes of the example 50R vial, the composite metric is the distance metric and the results of the summation process of block 954 on the values of Tables 6 and 7 are shown below in Table 8 and 9, respectively.Table 8 - non-defective composite values for each vialVIAL 1 READINGS VIAL 2 READINGS VIAL3 READINGSElectronically Filed May 6, 2026 DISTAN 100+101+102+103=406 101+102+103+104=410 102+103+104+105=414 CE SUMTable 9 - defect composite values for each vialVIAL 1 READINGS VIAL 2 READINGS VIAL 3 READINGS DISTAN 200+201+202+203= 201+202+203+204= 810 202+203+204+205=814 CE SUM 806

[0105] At block 956, the method 900 includes determining a value between the minimum defect composite value and the maximum non-defective composite value (such as an average or weighted average) to generate the composite metric alert threshold. In the case of the example values shown in Tables 8 and 9, execution of block 956 may average the minimum distance sum 806 from Table 9 with the maximum distance sum 414 from table 9 to generate a composite metric alert threshold of 610.

[0106] At block 958, the method 900 includes selecting the maximum value in the set of defective values for each other evaluation metric and defective test vial to generate a defect value candidate set for each other evaluation metric. The results of this selection on the values of Table 7 are shown below in Table 10.Table 10 - defect value candidate setVIAL 1 MAX VIAL 2 MAX VIAL 3 MAX READING READING READINGANGLE 134 135 136 MAXIMUMAREA MAXIMUM 59 60 60

[0107] At block 960, the method 900 includes selecting minimum value in the set of non-defective values for each other evaluation metric and non-defective test vial to generate a non-defective value candidate set for each other evaluation metric. The results of this selection on the values of Table 6 are shown below in Table 11.Electronically Filed May 6, 2026 Table 11 - non-defective value candidate setVIAL 1 VIAL 2 VIAL3 READINGS READINGS READINGSANGLE MINIMUM 129 130 131AREA MINIMUM 53 54 54

[0108] At block 960, the method 900 includes generating a secondary alert threshold by determining another value between the maximum value in the defect value candidate set with the maximum value in the non-defective value candidate set (such as by averaging or performing a weighted average) for the plurality of evaluation metrics. For example, execution of block 960 to the values in Tables 10 and 11 may include averaging the maximum angle 136 from Table 10 with the minimum angle 129 from Table 11 to generate a secondary alert threshold of 132.5 for the angle evaluation metric and averaging the maximum area 60 from Table 10 with the minimum area 53 from Table 11 to generate a secondary alert threshold of 56.5 for the area evaluation metric.

[0109] At block 962, the method 900 includes determining, for each of the defective and non-defective test vials, the number of instances where the representative set of defective values and the representative set of non-defective values (e.g., the values generated in block 952) exceeds the determined secondary alert threshold for each evaluation metric. The results of block 962 as applied to the values in Tables 6 and 7 are shown below in Tables 12 and 13, respectively.Table 12 - Instant count for non-defective vialsVIAL 1 VIAL 2 VIAL3 INSTANCES INSTANCES INSTANCES ANGLE 0 1 2AREA 0 1 1INSTANCES SUM 0 2 3Electronically Filed May 6, 2026 Table 13 - Instant count for defective vialsVIAL 1 VIAL 2 VIAL3 INSTANCES INSTANCES INSTANCES ANGLE 2 3 4AREA 3 4 4INSTANCES SUM 5 7 8

[0110] At block 964, the method 900 includes determining a value between the maximum number of instances identified for the non-defective test vials with the minimum number of instances for the defective test vials (such as by averaging or performing a weighted average) to generate an instance threshold to use with respect to evaluating the secondary alert thresholds. For example, execution of block 964 on the values in Tables 12 and 13 may average the maximum instance count of 3 from Table 12 with the minimum instance count of 5 from Table 13 to generate an instance threshold of 4.[oni] It should be appreciated that the values described above with respect to Tables 6-13 and blocks 952-965 are example values that illustrate the process of the method 950 with respect to example 50R type vials and the method 900 may be utilized to generate other values for similar or different types of vials, seal crimps, and evaluation metrics (e.g., the 20R vial shown in silhouette image 700 of FIG. 7 A).

[0112] Furthermore, in some embodiments, execution of the method 950 may produce the various multi-metric alert condition as described above. In one example, the composite alert metric is a distance sum and the composite metric alert threshold is a distance sum value greater than 36.6mm. In this example, a secondary alert threshold may be defined to additionally be based on an area metric threshold exceeding 1.50 mm2and an angle metric exceeding 150°. The multi-metric alert condition may also be associated with an instance threshold of 4 instances (e.g., the number of measured angles and areas that exceed the angle metric threshold and the areas metric threshold, respectively). The delineation of these thresholds with respect to sets of defective and non-defective values are shown in FIGs. 10A-10C. For example, FIG. 10A shows the set of defective vial angle values 1000 and the non-defective angle values 1002 relative to the angle secondary alert threshold 1003 as measured in images taken from both first camera 406A (e.g., “Head A”) and second camera 406B (e.g., “Head B”). FIG. 10B shows the set of defective vial area values 1004 and the non-defective area values 1006 relative to the area secondary alert threshold 1007. FIG. 10C shows the setElectronically Filed May 6, 2026 of defective vial distance sum values 1008 and the non-defective distance sum values 1010 relative to the distance sum composite metric alert threshold 1012.

[0113] Using these values, the multi-metric analysis would fail and reject the vial when at least four (e.g., the instant threshold value) of the eight angle and area measurements fail (e.g., the angle and area measurements taken from the silhouette images from both the first camera 406A and the second camera 406B) and the distance sum must fail (e.g, the sum of the distances measured in all four images exceeds 36.6mm) for the vial to be rejected. For example, if area evaluation metric 600A for the image from the first camera 406A measures 1.52 mm2(greater than 1.50 mm2) and the remaining area metrics measure 1.48 mm2(less than 1.50 mm2) and if angle evaluation metric 602A for the image from the first camera 406A measures 155° (greater than 150°) and the remaining angle metrics measure 145° (less than 1.50 mm2) the instances counter is 2 because only 2 measurement met the criteria. Thus, this vial will be accepted even if the distance sum exceeds 36.6 m. However, if both the area evaluation metrics 600A and 600B for the image from the first camera 406A measure 1.52 mm2(greater than 1.50 mm2) and, the second image area metrics measure 1.48 mm2(less than 1.50 mm2) and bot the angle evaluation metrics 602A and 602B measure 155° (greater than 150°) and the second image angle metrics measure 145° (less than 1.50 mm2) the instances counter is 4 because 4 measurements met the criteria. Therefore, if the distance sum of the four distance measurements (e.g, evaluation metrics 604A and 604B from both images) is 36.7mm (greater than 36.6mm), the vial will be rejected since the criteria of 4 instances and distance sum is met.

[0114] In some embodiments, a multi-metric alert condition includes a Boolean rule that combines a plurality of evaluation metrics to distinguish between non-defective and defective crimps. A Boolean rule can include a logical expression that combines multiple evaluation metrics using Boolean operators such as AND, OR, and NOT to produce a binary pass / fail outcome. A Boolean rule can also include arithmetic or statistical operations of the evaluation metrics (e.g., difference, standard deviation, average). A Boolean rule can be expressed as a disjunctive normal form expression. For example, a Boolean rule for a multi-metric alert condition may specify that a seal crimp is defective when (angle > thresholdl AND area > threshold2) OR (distance sum > thresholds AND instance count > threshold4). Such Boolean rules can enable efficient evaluation of complex multi-metric conditions while maintaining interpretability and computational efficiency suitable for high-throughput inspection environments.Electronically Filed May 6, 2026

[0115] In some embodiments, a Boolean rule for a multi-metric alert condition can be determined using an artificial intelligence (Al) algorithm (e.g., a rule miner). An Al rule miner can help identify multi-metric alerts conditions that otherwise might not be apparent. For example, an Al rule miner can receive a table of evaluation metrics extracted from inspection data (e.g., angles, distances, radii as described above) and labeled outcomes (defective / non-defective), then identify a compact Boolean rule for a multi-metric alert. The inspection data can be split for train / test / validation sets to train, test, and validate the Al rule miner. For example, an Al rule miner can receive as input feature tables of non-defective and defective units (e.g., as comma-separated values) and output a best-found decision (e.g., in TMX native language, typically a short disjunctive normal form (DNF) expression).

[0116] In some embodiments, a greedy DNF (Disjunctive Normal Form) algorithm can be used to determine a Boolean expression for a multi-metric alert condition. A greedy DNF algorithm can construct a Boolean rule in disjunctive normal form incrementally, making locally optimal (“greedy”) choices at each step. A greedy DNF algorithm can combine several fine-tuned parameters (e.g., angles, distances, radius, and asymmetry metrics) into a single, interpretable pass / fail rule that can be directly implemented in the advanced calculations as a set of AND / OR conditions. These unified decision expressions represent a trade-off between detection performance and rule complexity based on the available characterization data. The Boolean expressions determined using the greedy DNF algorithm may not be final or unique. For example, if future data, re-tuning, or simplification work identifies a more optimal or more compact expression that maintains or improves detection performance and false-fail rates, the criteria and corresponding Boolean expressions can be updated accordingly without changing the overall methodology.

[0117] For example, a greedy DNF algorithm can construct a Boolean rule incrementally by performing a bounded beam search over candidate Boolean clauses composed of threshold-based literals corresponding to evaluation metrics. Candidate literals can be generated using threshold values derived from data for defective and non-defective vials. This data can optionally be supplemented with threshold hints or seed expressions from auxiliary models. The greedy DNF algorithm can expand clauses incrementally, for example, by adding one literal at a time, optionally preventing repeated use of the same metric within a clause, and can rank candidate clauses according to one or more of training performance, validation performance, defective-unit recall, non-defective-unit false-fail rate, incremental capture of previously uncaptured defective units, overlap with previously selected clauses, and rule complexity. In multi-image inspection workflows, candidate clauses can beElectronically Filed May 6, 2026 evaluated in unit-level decision space by aggregating multiple images associated with a common vial according to a K-of-N criterion. The resulting Boolean expression can be simplified prior to deployment by removing redundant literals, absorbing redundant clauses, pruning on validation data, or adjusting thresholds to improve robustness.

[0118] In some embodiments, an Al rule miner can use one or more auxiliary search backends, including tree-based, ensemble, sparse linear, or reduced exact-search models, to generate feature rankings, candidate literals, threshold hints, or seed expressions for a greedy DNF search. In some embodiments, the Al rule miner can first apply an anomaly-prefix stage that identifies obvious defective units using high-sigma single-literal rules and conjunctive anomaly rules, and then perform the greedy DNF search on remaining residual data. For example, an Al rule miner can enforce hard constraints on false-positive and false-negative rates and / or can optimize rule selection according to balanced, Pareto, false-positive-priority, false-negative-priority, or cost-sensitive objectives.

[0119] In multi-image workflows where silhouette images are captured from multiple cameras, the methodology for determining multi-metric alert conditions can incorporate both per-image evaluation metrics and cross-image aggregation of evaluation metrics into the evaluation metric table. Per-image evaluation metrics can include edge strength, angle or radius analogs, intensity statistics, and local geometry descriptors. Cross-image aggregation evaluation metrics can include minimum, maximum, or mean values across images, within image, image-to-image asymmetry metrics, flags indicating conditions present in any image, and variability or standard deviation across views. Once these evaluation metrics are assembled into a single unit-level dataset, the Al rule miner can search for Boolean expression solutions that are both performant and deployable. For example, this approach can result in rules that reflect multi-image logic, such as failing a vial if any view shows a particular condition, failing if at least two views exceed a threshold, or failing if variability across views indicates inconsistency. The output remains an interpretable AND / OR threshold rule, but encodes multi-view evidence in a manner that aligns well with how inspection systems capture and evaluate inspection data.

[0120] In some embodiments, the defect detection system 404 can implement an automated vision module configured to design, execute, and refine the inspection pipeline end-to-end. In this configuration, the vision module can plan image-processing steps, execute those steps using machine-vision libraries, extract candidate evaluation metrics or features, and use the Al rule miner as an objective verifier to determine whether the extracted evaluation metric set can yield a compact, deployable Boolean rule with acceptableElectronically Filed May 6, 2026 performance. The vision module can form a closed loop where, if the Al rule miner cannot find an adequately accurate and compact rule, the vision module modifies the imageprocessing pipeline and iterates. Modifications to the image-processing pipeline can include adjustments to region of interest selection, filtering, denoising, edge detection, contour fitting, arc and radius extraction, and geometric stabilization. The Al rule miner can serve as an automated method to discover short Boolean decision expressions from measured evaluation metrics and to evaluate them via validation of metrics.

[0121] In some embodiments, the vision module can consider an alternative path where, instead of using an interpretable Boolean expression, the system can train a deep-learning model directly on the images or intermediate representations to produce the pass / fail decision. In that configuration, the Al rule miner can remain valuable as an interpretable benchmark and as a deployment-friendly option when a compact AND / OR rule is preferred, but the pipeline can switch to deep learning when the feature-to-rule route is insufficient for the desired separation between non-defective and defective classes.

[0122] To make the vision module effective in real inspection environments, the vision module can incorporate a memory or knowledge layer that captures how experienced vision engineers build, debug, and stabilize inspections. For example, the knowledge layer can include reusable recipes indicating what to try first and learned troubleshooting patterns indicating what to change when measurements drift or false fails increase. In another example, the memory or knowledge layer can store and reuse canonical pipeline templates, such as starting with locator or pattern alignment, defining regions of interest, performing edge extraction, fitting geometry, and computing derived metrics. In some embodiments, the memory or knowledge layer may also store failure-mode playbooks. For example, if edge instability occurs, the vision module can adjust region of interest placement, filtering, or thresholding. In another example, if asymmetry rises, the vision expert may add left and right comparisons. If borderline separation is poor, the vision expert may introduce additional angles, radius measurements, or derived statistics such as standard deviation.

[0123] In some embodiments, the memory or knowledge layer can further store rule and feature audit history so that the vision expert learns which evaluation metric families tend to yield compact, deployable Boolean expressions and which do not. The memory or knowledge layer can also incorporate closed-loop learning signals where the rule miner's success or failure and confusion-matrix performance become feedback signals that update the memory of what works for a given vial format, lighting condition, or station configuration. With this memory component, the vision module can learn a growing body of practical vision expertiseElectronically Filed May 6, 2026 including what to try, in what order, and why, while still using the Al rule miner as an objective verifier and deployment pathway when a compact Boolean decision expression is achievable.

[0124] FIG. 11 is a flowchart of an example method 1100 for identifying a defective seal crimp. The method 1100 may be executed by the defect detection system 404 of FIG. 4 or another suitable computing system.

[0125] At block 1110, the method 1100 includes generating a silhouette image (e.g., silhouette image 500 or silhouette image 700) of a vial (e.g., sealed vial 100, vial 105, vial 205a, etc.) located between a camera (e.g., first camera 406A or second camera 406B) and a backlight (e.g., first backlight 408 A or second backlight 408B) at a defect detection station (e.g., defect detection station 400) of an assembly line system. The vial may have an associated crimp type. The camera may comprise a telecentric lens. The backlight may be telecentrically arranged with respect to the camera.

[0126] At block 1120, the method 1100 includes identifying, via one or more processors (e.g., processing unit 422 of control unit 410), a plurality of zones (e.g., zones 506A, 506B, 508A, 508B, 510A, and 510B) within the silhouette image. The plurality of zones are defined based on a reference pattern (e.g., reference pattern 504) associated with the crimp type.

[0127] The method 1100 may include defining the reference pattern associated with the crimp type by generating a representative silhouette image of a crimp defect free vial having the crimp type; defining, via the one or more processors, a size and relative location of the plurality of zones within the representative silhouette image; and saving, via the one or more processors, the defined zones as the reference pattern associated with the crimp type. The method 1100 may also include defining, via the one or more processors, the plurality of evaluation metrics with respect to the plurality of zones; and associating, via the one or more processors, the defined plurality of evaluation metrics with the reference pattern.

[0128] At block 1130, the method 1100 includes analyzing, via the one or more processors, the silhouette image to measure a plurality of evaluation metrics (e.g., area evaluation metrics 600A and 600B, angle evaluation metrics 602A and 602B, distance evaluation metrics 604A and 604B, radius evaluation metrics 605A and 605B, intersection distance evaluation metrics 606A1 and 606B1, angle 708 A, and / or angle 708B, ) for a seal crimp of the vial. The measurements are based on image data within the identified plurality of zones.

[0129] At block 1140, the method 1100 includes determining, via the one or more processors, whether the seal crimp of the vial includes a defect by determining whether atElectronically Filed May 6, 2026 least one of the plurality of evaluation metrics exhibits a corresponding alert condition.

[0130] The method 1100 may include defining the alert conditions by: generating, via the one or more processors, a representative set of defective values for each evaluation metric and a representative set of non-defective values for each evaluation metric; and defining, via the one or more processors, the alert conditions for the plurality of evaluation metrics as a threshold value that delineates the representative set of defective values and the representative set of non-defective values for each of the evaluation metrics. The method 1100 may also include generating the representative set of defective values and the representative set of non-defective values by: generating a first set of testing silhouette images for defect free vials having the crimp type; generating a second set of testing silhouette images for defective crimp vials having the crimp type; identifying, via the one or more processors, the plurality of zones within the first set of silhouette images and the second set of silhouette images based on the reference pattern associated with the crimp type; and measuring, within the first set of silhouette images and the second set of silhouette images via the one or more processors, the plurality of evaluation metrics for seal crimps of the defect free vials and the defective crimp vials with respect to the plurality of zones to generate the representative set of defective values and the representative set of non-defective values. The alert conditions may include a multi-metric alert condition for evaluating the plurality of evaluation metrics. Defining the multi-metric alert condition may include defining, via the one or more processors, respective secondary threshold values for the plurality of evaluation metrics that also delineate the representative sets of defective values and the representative sets of non-defective values for each of the evaluation metrics. The multi-metric alert condition is exhibited when two of the measurements of the plurality of evaluation metrics exceed the respective secondary threshold values. Defining the multi-metric alert condition can include identifying a Boolean rule that combines a plurality of evaluation metrics. A Boolean rule can be determined using an artificial intelligence algorithm (e.g., a rule miner)

[0131] At block 1150, the method 1100 includes rejecting or accepting the vial based on whether the seal crimp of the vial includes the defect. Rejecting the vial may include directing the vial to a defective vial location of the assembly line system.

[0132] In some embodiments, the plurality of zones include at least one of: a top zone referenced to a top portion of a flip-off portion of the seal crimp; a side zone referenced to a side portion of the seal crimp; and a bottom zone referenced to a bottom portion of the seal crimp. In some embodiments, the plurality of evaluation metrics include an angle between a first edge location (e.g., edge locations 514A or 514B) within the side zone and a secondElectronically Filed May 6, 2026 edge location (e.g., edge locations 516A or 516B) within the bottom zone. The method 1100 may also include measuring the angle by: scanning, via the one or more processors, the side zone for changes in pixel values that identify the first edge; scanning, via the one or more processors, the bottom zone for changes in pixel values that identify the second edge; and measuring, via the one or more processors, the angle between the first edge and the second edge. In some embodiments, the plurality of evaluation metrics include a plurality of angles, each angle defined between a first edge location within a side zone (e.g., edge locations 514A or 514B) and one of a plurality of edge locations within the bottom zone (516A1, 516A2, 516A3, 516A4, 516B1, 516B2, 516B3, 516B4). In some embodiments, the plurality of evaluation metrics may also include an area (e.g., area evaluation metrics 600A or 600B) within the bottom zone. The method 1100 may include measuring the area within the one of the plurality of zones by: measuring, via the one or more processors, the area of pixels within a silhouette (e.g., silhouette 502) of the vial located within the bottom zone. In some embodiments, the plurality of evaluation metrics include a radius of an arc fitted to a first edge location within the side zone (e.g., edge locations 514A or 514B) and a second edge location within the bottom zone (e.g., edge locations 516A or 516B). In some embodiments, the plurality of evaluation metrics include a distance measured from a first edge location (e.g., edge locations 514A or 514B) to an intersection of the first edge location within the side zone and a second edge location within the bottom zone (e.g., edge locations 516A or 516B) along a defined direction. In some embodiments, the plurality of evaluation metrics include an intersection distance measured from one of a plurality of crimp edge locations (e.g., edge locations 516A1, 516A2, 516A3, 516A4, 516B1, 516B2, 516B3, 516B4) and an intersection (e.g., intersection 520A, 520B) of a first edge location within a side zone (e.g., edge locations 514A, 514B) and a second edge location within the bottom zone (e.g., edge locations 518A, 518A). In some embodiments, intersection distance metrics can be measured at a plurality of crimp edge locations.

[0133] In some embodiments, the side zone is a first side zone and the bottom zone is a first bottom zone; the plurality of zones include a second side zone referenced to another side portion of the seal crimp and a second bottom zone referenced to another bottom portion of the seal crimp; the plurality of evaluation metrics include a difference between a first angle and a second angle; the first angle is between a first edge location in the first side zone and a second edge location in the first bottom zone; and the second angle is between a third edge location in the second side zone and a fourth edge location in the second bottom zone. In these embodiments, the method 1100 may include determining the difference by: scanning,Electronically Filed May 6, 2026 via the one or more processors, the first side zone for changes in pixel values that identify the first edge; scanning, via the one or more processors, the first bottom zone for changes in pixel values that identify the second edge; scanning, via the one or more processors, the second side zone for changes in pixel values that identify the third edge; scanning, via the one or more processors, the second bottom zone for changes in pixel values that identify the fourth edge; measuring, via the one or more processors, the first angle between the first edge and the second edge and the second angle between the third edge and the fourth edge; and subtracting, via the one or more processors, the first angle from the second angle to determine the difference.

[0134] The method 1100 may also include determining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that the silhouette image fails to match the reference pattern associated with the crimp type. The silhouette image may fail to match the reference pattern associated with the crimp type when the plurality of zones fail to be identified in the silhouette image. The defect is one or more of a missing stopper, a missing cap, or a missing flip-off.

[0135] In some embodiments, the silhouette image is a first silhouette image, the camera is a first camera (e.g., first camera 406 A), and the backlight is a first backlight (e.g., first backlight 408 A). In these embodiments, the method 1100 may include generating a second silhouette image of the vial located between a second camera (e.g., second camera 406B) and a second backlight (e.g., second backlight 408B) at the defect detection station of the assembly line system; identifying, via one or more processors, the plurality of zones within the second silhouette image based on the reference pattern associated with the crimp type; analyzing, via the one or more processors, the second silhouette image to measure the plurality of evaluation metrics for the seal crimp of the vial with respect to the plurality of zones; and determining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that at least one of the plurality evaluation metrics measured from the second silhouette image exhibits the corresponding alert condition. The first camera and the second camera may be mounted to a single bracket. The first camera and the second camera may be oriented relative to each other at an angle in a range of 65-95 degrees. In some embodiments, the method 1100 may include averaging one of the plurality evaluation metrics measured from the first silhouette image with a corresponding one of the plurality evaluation metrics measured from the second silhouette image to generate an average metric; and determining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that the average metric exhibits the correspondingElectronically Filed May 6, 2026 alert condition.

[0136] In some embodiments, the plurality of zones include a right top zone referenced to a top right portion of a flip-off portion of the seal crimp; a left top zone referenced to a top left portion of the flip-off portion of the seal crimp; a right side zone referenced to a right side portion of the seal crimp; a left side zone referenced to a left side portion of the seal crimp; a right bottom zone referenced to a bottom right portion of the seal crimp. In these embodiments, the plurality of evaluation metrics include: a right side angle between a right edge location in the right side zone and a first right bottom edge location in the right bottom zone; a left side angle between a left edge location in the left side zone and a first left bottom edge location in the left bottom zone; a right side distance between a top edge location of the right top zone and a second bottom edge location in the right bottom zone; a left side distance between a top edge location of the left top zone and a second bottom edge location in the left bottom zone; a bottom right area within the right bottom zone; and a bottom left area within the left bottom zone.

[0137] In some embodiments, the method 1100 includes : determining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that (i) more than one of the right side angle, the left side angle, the bottom right area, and the bottom left area as measured in either the first silhouette image or the second silhouette image exceed a secondary threshold and (ii) a sum of the left side distance and right side distance measured from the first silhouette image and the left side distance and right side distance measured from the second silhouette image exceeds a composite distance threshold.

[0138] Without limitation, some example embodiments / features of the present disclosure include:

[0139] El. A method comprising: generating a silhouette image of a vial located between a camera and a backlight at a defect detection station of an assembly line system, the vial having an associated crimp type; identifying, via one or more processors, a plurality of zones within the silhouette image, wherein the plurality of zones are defined based on a reference pattern associated with the crimp type; analyzing, via the one or more processors, the silhouette image to measure a plurality of evaluation metrics for a seal crimp of the vial, wherein the measurements are based on image data within the identified plurality of zones; determining, via the one or more processors, whether the seal crimp of the vial includes a defect by determining whether at least one of the plurality of evaluation metrics exhibits a corresponding alert condition; and rejecting or accepting the vial by determining whether the seal crimp of the vial includes the defect.Electronically Filed May 6, 2026

[0140] E2. The method of El, further comprising defining the reference pattern associated with the crimp type by: generating a representative silhouette image of a crimp defect free vial having the crimp type; defining, via the one or more processors, a size and relative location of the plurality of zones within the representative silhouette image; and saving, via the one or more processors, the defined zones as the reference pattern associated with the crimp type.

[0141] E3. The method of El, further comprising: defining, via the one or more processors, the plurality of evaluation metrics with respect to the plurality of zones; and associating, via the one or more processors, the defined plurality of evaluation metrics with the reference pattern.

[0142] E4. The method of any of E1-E3, further comprising defining the alert conditions by: generating, via the one or more processors, a representative set of defective values for each evaluation metric and a representative set of non-defective values for each evaluation metric; and defining, via the one or more processors, respective primary threshold values that delineate the representative sets of defective values and the representative sets of nondefective values for each of the evaluation metrics, wherein the alert conditions include the plurality of evaluation metrics exceeding the respective primary threshold values.

[0143] E5. The method of E4, further comprising generating the representative set of defective values and the representative set of non-defective values by: generating a first set of testing silhouette images for defect free vials having the crimp type; generating a second set of testing silhouette images for defective crimp vials having the crimp type; identifying, via the one or more processors, the plurality of zones within the first set of silhouette images and the second set of silhouette images based on the reference pattern associated with the crimp type; and measuring, within the first set of silhouette images and the second set of silhouette images via the one or more processors, the plurality of evaluation metrics for seal crimps of the defect free vials and the defective crimp vials with respect to the plurality of zones to generate the representative set of defective values and the representative set of non-defective values.

[0144] E6. The method of E4, wherein: the alert conditions include a multi -metric alert condition for evaluating the plurality of evaluation metrics; defining the multi-metric alert condition includes defining, via the one or more processors, respective secondary threshold values for the plurality of evaluation metrics that also delineate the representative sets of defective values and the representative sets of non-defective values for each of the evaluationElectronically Filed May 6, 2026 metrics; and the multi-metric alert condition is exhibited when two of the measurements of the plurality of evaluation metrics exceed the respective secondary threshold values.

[0145] E7. The method of any of E1-E6, wherein the camera comprises a telecentric lens.

[0146] E8. The method of E7, wherein the backlight is telecentrically arranged with respect to the camera.

[0147] E9. The method of any of E1-E8, wherein the plurality of zones include at least one of: a top zone referenced to a top portion of a flip-off portion of the seal crimp; a side zone referenced to a side portion of the seal crimp; and a bottom zone referenced to a bottom portion of the seal crimp.

[0148] E10. The method of E9, wherein: the plurality of evaluation metrics include an angle between a first edge location within the side zone and a second edge location within the bottom zone; and the method further comprises measuring the angle by: scanning, via the one or more processors, the side zone for changes in pixel values that identify the first edge; scanning, via the one or more processors, the bottom zone for changes in pixel values that identify the second edge; and measuring, via the one or more processors, the angle between the first edge and the second edge.

[0149] Ell. The method of E9, wherein: the plurality of evaluation metrics include an area within the bottom zone; and the method further comprises measuring the area within the one of the plurality of zones by: measuring, via the one or more processors, the area of pixels within a silhouette of the vial located within the bottom zone.

[0150] E12. The method of E9, wherein: the side zone is a first side zone and the bottom zone is a first bottom zone; the plurality of zones include a second side zone referenced to another side portion of the seal crimp and a second bottom zone referenced to another bottom portion of the seal crimp; the plurality of evaluation metrics include a difference between a first angle and a second angle; the first angle is between a first edge location in the first side zone and a second edge location in the first bottom zone; and the second angle is between a third edge location in the second side zone and a fourth edge location in the second bottom zone.

[0151] El 3. The method of El 2, further comprising determining the difference by: scanning, via the one or more processors, the first side zone for changes in pixel values that identify the first edge; scanning, via the one or more processors, the first bottom zone for changes in pixel values that identify the second edge; scanning, via the one or more processors, the second side zone for changes in pixel values that identify the third edge;Electronically Filed May 6, 2026 scanning, via the one or more processors, the second bottom zone for changes in pixel values that identify the fourth edge; measuring, via the one or more processors, the first angle between the first edge and the second edge and the second angle between the third edge and the fourth edge; and subtracting, via the one or more processors, the first angle from the second angle to determine the difference.

[0152] E14. The method of any of E1-E13, further comprising: determining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that the silhouette image fails to match the reference pattern associated with the crimp type.

[0153] E15. The method of E14, wherein the silhouette image fails to match the reference pattern associated with the crimp type when the plurality of zones fail to be identified in the silhouette image.

[0154] El 6. The method of any of El -El 5, wherein the defect is one or more of a missing stopper, a missing cap, or a missing flip-off.

[0155] El 7. The method of any of El -El 6, wherein the silhouette image is a first silhouette image, the camera is a first camera, and the backlight is a first backlight and further comprising: generating a second silhouette image of the vial located between a second camera and a second backlight at the defect detection station of the assembly line system; identifying, via one or more processors, the plurality of zones within the second silhouette image based on the reference pattern associated with the crimp type; analyzing, via the one or more processors, the second silhouette image to measure the plurality of evaluation metrics for the seal crimp of the vial with respect to the plurality of zones; and determining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that at least one of the plurality evaluation metrics measured from the second silhouette image exhibits the corresponding alert condition.

[0156] El 8. The method of El 7, wherein the first camera and the second camera are mounted to a single bracket.

[0157] El 9. The method of El 7, wherein the first camera and the second camera are oriented relative to each other at an angle in a range of 65-95 degrees.

[0158] E20. The method of El 7, further comprising: averaging one of the plurality evaluation metrics measured from the first silhouette image with a corresponding one of the plurality evaluation metrics measured from the second silhouette image to generate an average metric; and determining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that the average metric exhibits the corresponding alert condition.Electronically Filed May 6, 2026

[0159] E21. The method of El 7 wherein: the plurality of zones include: a right top zone referenced to a top right portion of a flip-off portion of the seal crimp; a left top zone referenced to a top left portion of the flip-off portion of the seal crimp; a right side zone referenced to a right side portion of the seal crimp; a left side zone referenced to a left side portion of the seal crimp; and a right bottom zone referenced to a bottom right portion of the seal crimp; and the plurality of evaluation metrics include: a right side angle between a right edge location in the right side zone and a first right bottom edge location in the right bottom zone; a left side angle between a left edge location in the left side zone and a first left bottom edge location in the left bottom zone; a right side distance between a top edge location of the right top zone and a second bottom edge location in the right bottom zone; a left side distance between a top edge location of the left top zone and a second bottom edge location in the left bottom zone; a bottom right area within the right bottom zone; and a bottom left area within the left bottom zone.

[0160] E22. The method of E21, further comprising: determining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that (i) more than one of the right side angle, the left side angle, the bottom right area, and the bottom left area as measured in either the first silhouette image or the second silhouette image exceed a secondary threshold and (ii) a sum of the left side distance and right side distance measured from the first silhouette image and the left side distance and right side distance measured from the second silhouette image exceeds a composite distance threshold.

[0161] E23. The method of any of E1-E22, wherein rejecting the vial includes directing the vial to a defective vial location of the assembly line system.

[0162] E24. A defect detection system comprising: at least one camera; at least one backlight arranged telecentrically with the at least one camera; a memory storing a reference pattern associated with a vial crimp type; and a control unit having one or more processors operatively connected with the at least one camera and the memory and one or more non-transitory, computer-readable media storing instructions that, when executed by the one or more processors, cause the control unit to: receive a silhouette image of a vial generated by the at least one camera while the vial is located between the at least one camera and the at least one backlight, the vial having the vial crimp type; identify a plurality of zones within the silhouette image, wherein the plurality of zones are defined based on a reference pattern associated with the crimp type; analyze the silhouette image to measure a plurality of evaluation metrics for a seal crimp of the vial, wherein the measurements are based on image data within the identified plurality of zones; determine that the seal crimp of the vial includesElectronically Filed May 6, 2026 a defect by determining that at least one of the plurality of evaluation metrics exhibits a corresponding alert condition; and generate an alert signal indicating that the vial is defective by determining that the seal crimp of the vial includes the defect.

[0163] E25. The defect detection system of E24, wherein the pattern associated with the crimp type is defined by: generating a representative silhouette image of a crimp defect free vial having the crimp type; defining a size and relative location of the plurality of zones within the representative silhouette image; and saving the defined zones as the reference pattern associated with the crimp type.

[0164] E26. The defect detection system of E24, wherein the reference pattern associated with the crimp type is defined by: defining the plurality of evaluation metrics with respect to the plurality of zones; and associating the defined plurality of evaluation metrics with the reference pattern.

[0165] E27. The defect detection system of any of E24-E26, wherein the alert conditions are defined by: generating a representative set of defective values for each evaluation metric and a representative set of non-defective values for each evaluation metric; and defining respective primary threshold values that delineate the representative sets of defective values and the representative sets of non-defective values for each of the evaluation metrics, wherein the alert conditions include the plurality of evaluation metrics exceeding the respective primary threshold values.

[0166] E28. The defect detection system of E27, wherein generating the representative set of defective values and the representative set of non-defective values includes: generating a first set of testing silhouette images for defect free vials having the crimp type; generating a second set of testing silhouette images for defective crimp vials having the crimp type; identifying the plurality of zones within the first set of silhouette images and the second set of silhouette images based on the reference pattern associated with the crimp type; and measuring, within the first set of silhouette images and the second set of silhouette images, the plurality of evaluation metrics for seal crimps of the defect free vials and the defective crimp vials with respect to the plurality of zones to generate the representative set of defective values and the representative set of non-defective values.

[0167] E29. The defect detection system of E27, wherein: the alert conditions include a multi-metric alert condition for evaluating the plurality of evaluation metrics; defining the multi-metric alert condition includes defining, via the one or more processors, respective secondary threshold values for the plurality of evaluation metrics that also delineate the representative sets of defective values and the representative sets of non-defective values forElectronically Filed May 6, 2026 each of the evaluation metrics; and the multi-metric alert condition is exhibited when two of the measurements of the plurality of evaluation metrics exceed the respective secondary threshold values.

[0168] E30. The defect detection system of any of E24-E29, wherein the at least one camera comprises a telecentric lens.

[0169] E31. The defect detection system of any of E24-E30, wherein the plurality of zones include at least one of: a top zone referenced to a top portion of a flip-off portion of the seal crimp; a side zone referenced to a side portion of the seal crimp; and a bottom zone referenced to a bottom portion of the seal crimp.

[0170] E32. The defect detection system of E31, wherein the plurality of evaluation metrics include an angle between a first edge location within the side zone and a second edge location within the bottom zone and the angle is measured by: scanning, the side zone for changes in pixel values that identify the first edge; scanning, the bottom zone for changes in pixel values that identify the second edge; and measuring the angle between the first edge and the second edge.

[0171] E33. The defect detection system of E31, wherein the plurality of evaluation metrics include an area within the bottom zone and the area within the one of the plurality of zones is measured by: measuring the area of pixels within a silhouette of the vial located within the bottom zone.

[0172] E34. The defect detection system of E31, wherein: the side zone is a first side zone and the bottom zone is a first bottom zone; the plurality of zones include a second side zone referenced to another side portion of the seal crimp and a second bottom zone referenced to another bottom portion of the seal crimp; the plurality of evaluation metrics include a difference between a first angle and a second angle; the first angle is between a first edge location in the first side zone and a second edge location in the first bottom zone; and the second angle is between a third edge location in the second side zone and a fourth edge location in the second bottom zone.

[0173] E35. The defect detection system of E34, wherein the difference is determined by: scanning the first side zone for changes in pixel values that identify the first edge; scanning the first bottom zone for changes in pixel values that identify the second edge; scanning the second side zone for changes in pixel values that identify the third edge; scanning the second bottom zone for changes in pixel values that identify the fourth edge; measuring the first angle between the first edge and the second edge and the second angleElectronically Filed May 6, 2026 between the third edge and the fourth edge; and subtracting the first angle from the second angle to determine the difference.

[0174] E36. The defect detection system of any of E24-E35, wherein the instructions, when executed by the one or more processors, cause the control unit to: determine that the seal crimp of the vial includes the defect by determining that the silhouette image fails to match the reference pattern associated with the crimp type.

[0175] E37. The defect detection system of E36, wherein the silhouette image fails to match the reference pattern associated with the crimp type when the one or more processors fail to identify the plurality of zones in the silhouette image.

[0176] E38. The defect detection system of any of E24-E37, wherein the defect is one or more of a missing stopper, a missing cap, or a missing flip-off.

[0177] E39. The defect detection system of any of E24-E38, wherein the silhouette image is a first silhouette image, the one or more cameras are a first camera, and the one or more backlights are a first backlight and further comprising: a second camera; and a second backlight arranged telecentrically with the second camera, wherein, the instructions, when executed by the one or more processors, cause the control unit to: receive a second silhouette image of the vial the vial located between the second camera and the second backlight; identify the plurality of zones within the second silhouette image based on the reference pattern associated with the crimp type; analyze the second silhouette image to measure the plurality of evaluation metrics for the seal crimp of the vial with respect to the plurality of zones; and determine that the seal crimp of the vial includes the defect by determining that at least one of the plurality evaluation metrics measured from the second silhouette image exhibits the corresponding alert condition.

[0178] E40. The defect detection system of E39, wherein the first camera and the second camera are mounted to a single bracket.

[0179] E41. The defect detection system of E39, wherein the first camera and the second camera are oriented relative to each other at an angle in a range of 65-95 degrees.

[0180] E42. The defect detection system of E39, wherein the instructions, when executed by the one or more processors, cause the control unit to: average one of the plurality evaluation metrics measured from the first silhouette image with a corresponding one of the plurality evaluation metrics measured from the second silhouette image to generate an average metric; and determine that the seal crimp of the vial includes the defect by determining that the average metric exhibits the corresponding alert condition.Electronically Filed May 6, 2026

[0181] E43. The defect detection system of E39 wherein: the plurality of zones include: a right top zone referenced to a top right portion of a flip-off portion of the seal crimp; a left top zone referenced to a top left portion of the flip-off portion of the seal crimp; a right side zone referenced to a right side portion of the seal crimp; a left side zone referenced to a left side portion of the seal crimp; and a right bottom zone referenced to a bottom right portion of the seal crimp; and the plurality of evaluation metrics include: a right side angle between a right edge location in the right side zone and a first right bottom edge location in the right bottom zone; a left side angle between a left edge location in the left side zone and a first left bottom edge location in the left bottom zone; a right side distance between a top edge location of the right top zone and a second bottom edge location in the right bottom zone; a left side distance between a top edge location of the left top zone and a second bottom edge location in the left bottom zone; a bottom right area within the right bottom zone; and a bottom left area within the left bottom zone.

[0182] E44. The defect detection system of E43, wherein the instructions, when executed by the one or more processors, cause the control unit to: determine, that the seal crimp of the vial includes the defect by determining that (i) more than one of the right side angle, the left side angle, the bottom right area, and the bottom left area as measured in either the first silhouette image or the second silhouette image exceed a secondary threshold and (ii) a sum of the left side distance and right side distance measured from the first silhouette image and the left side distance and right side distance measured from the second silhouette image exceeds a composite distance threshold.

[0183] E45. An assembly line system comprising: the silhouette imaging and defect detection system of any of E24-E44 deployed at a defect detection station; and an assembly line controller configured to reject the vial by receiving the alert signal from the defect detection system.

[0184] E46. The assembly line system of E45, wherein to reject the vial assembly line controller directs the vial to a defective vial location of the assembly line system.

[0185] E47. A method comprising: generating a silhouette image of a vial located between a camera and a backlight at a defect detection station of an assembly line system, the vial having an associated crimp type; determining, via the one or more processors, whether the silhouette image matches a reference pattern associated with the crimp type, the reference pattern defining a plurality of zones to be located within the silhouette image; determining, via the one or more processors, whether the seal crimp of the vial includes a defect by determining whether the silhouette image fails to match the reference pattern associated withElectronically Filed May 6, 2026 the crimp type; and rejecting or accepting the vial by determining whether the seal crimp of the vial includes the defect.

[0186] E48. The method of E47, wherein the silhouette image fails to match the reference pattern associated with the crimp type when the plurality of zones fail to be identified in the silhouette image.

[0187] E49. The method of E47 or E48, further comprising: identifying, via the one or more processors, the plurality of zones within the silhouette image; analyzing, via the one or more processors, the silhouette image to measure a plurality of evaluation metrics for the seal crimp of the vial, wherein the measurements are based on image data within the identified plurality of zones; and determining, via the one or more processors, whether the seal crimp of the vial includes the defect by determining whether at least one of the plurality of evaluation metrics exhibits a corresponding alert condition.

[0188] E50. A defect detection system comprising: at least one camera; at least one backlight arranged telecentrically with the at least one camera; a memory storing a reference pattern associated with a vial crimp type; and a control unit having one or more processors operatively connected with the at least one camera and the memory and one or more non-transitory, computer-readable media storing instructions that, when executed by the one or more processors, cause the control unit to: receive a silhouette image of a vial generated by the at least one camera while the vial is located between the at least one camera and the at least one backlight, the vial having the vial crimp type; determine whether the silhouette image matches a reference pattern associated with the crimp type, the reference pattern defining a plurality of zones to be located within the silhouette image; determine whether the seal crimp of the vial includes a defect by determining whether the silhouette image fails to match the reference pattern associated with the crimp type; and generate an alert signal indicating that the vial is defective by determining that the seal crimp of the vial includes the defect.

[0189] E51. The defect detection system of E50, wherein the silhouette image fails to match the reference pattern associated with the crimp type when the plurality of zones fail to be identified in the silhouette image.

[0190] E52. The defect detection system of E50 or E51, wherein the instructions, when executed by the one or more processors, cause the control unit to: identify the plurality of zones within the silhouette image; analyze the silhouette image to measure a plurality of evaluation metrics for the seal crimp of the vial, wherein the measurements are based on image data within the identified plurality of zones; and determine whether the seal crimp ofElectronically Filed May 6, 2026 the vial includes the defect by determining whether at least one of the plurality of evaluation metrics exhibits a corresponding alert condition.

[0191] Without limitation, some additional example embodiments / features of the present disclosure include:

[0192] E53. A method comprising: generating a silhouette image of a vial located between a camera and a backlight at a defect detection station of an assembly line system, the vial having an associated crimp type; identifying, via one or more processors, a plurality of zones within the silhouette image, wherein the plurality of zones are defined based on a reference pattern associated with the crimp type; analyzing, via the one or more processors, the silhouette image to measure a plurality of evaluation metrics for a seal crimp of the vial, wherein the measurements are based on image data within the identified plurality of zones; determining, via the one or more processors, whether the seal crimp of the vial includes a defect by determining whether at least one of the plurality of evaluation metrics exhibits a corresponding alert condition; and rejecting or accepting the vial based on whether the seal crimp of the vial includes the defect.

[0193] E54. The method of E53, further comprising defining the reference pattern associated with the crimp type by: generating a representative silhouette image of a crimp defect free vial having the crimp type; defining, via the one or more processors, a size and relative location of the plurality of zones within the representative silhouette image; and saving, via the one or more processors, the defined zones as the reference pattern associated with the crimp type.

[0194] E55. The method of E53 or E54, further comprising: defining, via the one or more processors, the plurality of evaluation metrics with respect to the plurality of zones; and associating, via the one or more processors, the defined plurality of evaluation metrics with the reference pattern.

[0195] E56. The method of any of E53-E55, further comprising defining the alert condition by: defining, via the one or more processors, respective primary threshold values for each of the evaluation metrics, wherein the alert conditions include one or more of the plurality of evaluation metrics outside the respective primary threshold values.

[0196] E57. The method of any of E53-E56, further comprising defining the alert condition by: generating, via the one or more processors, a representative set of defective values for each evaluation metric and a representative set of non-defective values for each evaluation metric; and defining, via the one or more processors, respective primary threshold values that delineate the representative sets of defective values and the representative sets ofElectronically Filed May 6, 2026 non-defective values for each of the evaluation metrics, wherein the alert conditions include one or more of the plurality of evaluation metrics outside the respective primary threshold values.

[0197] E58. The method of E57, further comprising generating the representative set of defective values for each evaluation metric and the representative set of non-defective values for each evaluation metric by: generating a first set of testing silhouette images for defect free vials having the crimp type; generating a second set of testing silhouette images for defective crimp vials having the crimp type; identifying, via the one or more processors, the plurality of zones within the first set of silhouette images and the second set of silhouette images based on the reference pattern associated with the crimp type; and measuring, within the first set of silhouette images and the second set of silhouette images via the one or more processors, the plurality of evaluation metrics for seal crimps of the defect free vials and the defective crimp vials with respect to the plurality of zones to generate the representative set of defective values and the representative set of non-defective values.

[0198] E59. The method of any of E53-E58, wherein the alert condition includes a multimetric alert condition.

[0199] E60. The method of E59, wherein the multi-metric alert condition includes a Boolean rule.

[0200] E61. The method of E60, wherein the Boolean rule is determined using a greedy disjunctive normal form algorithm.

[0201] E62. The method of E60 or E61, wherein the Boolean rule is determined using an artificial intelligence algorithm.

[0202] E63. The method of any of E59-E62, wherein the multi -metric alert condition is defined by defining, via the one or more processors, respective secondary threshold values for the plurality of evaluation metrics, and the multi-metric alert condition is exhibited when two or more of the measurements of the plurality of evaluation metrics are outside the respective secondary threshold values.

[0203] E64. The method of any of E57-E63, wherein: the alert condition includes a multi-metric alert condition for evaluating the plurality of evaluation metrics; the multimetric alert condition is defined by defining, via the one or more processors, respective secondary threshold values for the plurality of evaluation metrics that also delineate the representative sets of defective values and the representative sets of non-defective values for each of the evaluation metrics; and the multi-metric alert condition is exhibited when two ofElectronically Filed May 6, 2026 the measurements of the plurality of evaluation metrics are outside the respective secondary threshold values.

[0204] E65. The method of any of E53-E64, wherein the camera comprises a telecentric lens.

[0205] E66. The method of E65, wherein the backlight is telecentrically arranged with respect to the camera.

[0206] E67. The method of any of E53-E66, wherein the plurality of zones include at least one of: a top zone referenced to a top portion of a flip-off portion of the seal crimp; a side zone referenced to a side portion of the seal crimp; and a bottom zone referenced to a bottom portion of the seal crimp.

[0207] E68. The method of E67, wherein the plurality of evaluation metrics include an angle between a first edge location within the side zone and a second edge location within the bottom zone.

[0208] E69. The method of E67 or E68, wherein the plurality of evaluation metrics include a plurality of angles, each angle defined between a first edge location within a side zone and one of a plurality of edge locations within the bottom zone.

[0209] E70. The method of any of E67-E69, wherein: the plurality of evaluation metrics include an angle between a first edge location within the side zone and a second edge location within the bottom zone; and the method further comprises measuring the angle by: scanning, via the one or more processors, the side zone for changes in pixel values that identify the first edge; scanning, via the one or more processors, the bottom zone for changes in pixel values that identify the second edge; and measuring, via the one or more processors, the angle between the first edge and the second edge.

[0210] E71. The method of any of E67-E70, wherein the plurality of evaluation metrics include an area within the bottom zone.

[0211] E72. The method of any of E67-E71, wherein: the plurality of evaluation metrics include an area within the bottom zone; and the method further comprises measuring the area within bottom zone by: measuring, via the one or more processors, the area of pixels within a silhouette of the vial located within the bottom zone.

[0212] E73. The method of any of E67-E72, wherein the plurality of evaluation metrics include a radius of an arc fitted to a first edge location within the side zone and a second edge location within the bottom zone.Electronically Filed May 6, 2026

[0213] E74. The method of any of E67-E73, wherein the plurality of evaluation metrics include a distance measured between a crimp edge location and an intersection of a first edge location within the side zone and a second edge location within the bottom zone.

[0214] E75. The method of any of E67-E74, wherein the plurality of evaluation metrics include a plurality of intersection distance metrics at a plurality of crimp edge locations, each intersection distance metric measured between a one of the plurality of crimp edge locations and an intersection of a first edge location within the side zone and a second edge location within the bottom zone.

[0215] E76. The method of any of E67-E75, wherein: the side zone is a first side zone and the bottom zone is a first bottom zone; the plurality of zones include a second side zone referenced to another side portion of the seal crimp and a second bottom zone referenced to another bottom portion of the seal crimp; the plurality of evaluation metrics include a difference between a first angle and a second angle; the first angle is between a first edge location in the first side zone and a second edge location in the first bottom zone; and the second angle is between a third edge location in the second side zone and a fourth edge location in the second bottom zone.

[0216] E77. The method of E76, further comprising determining the difference by: scanning, via the one or more processors, the first side zone for changes in pixel values that identify the first edge; scanning, via the one or more processors, the first bottom zone for changes in pixel values that identify the second edge; scanning, via the one or more processors, the second side zone for changes in pixel values that identify the third edge; scanning, via the one or more processors, the second bottom zone for changes in pixel values that identify the fourth edge; measuring, via the one or more processors, the first angle between the first edge and the second edge and the second angle between the third edge and the fourth edge; and subtracting, via the one or more processors, the first angle from the second angle to determine the difference.

[0217] E78. The method of any of E53-E77, further comprising: determining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that the silhouette image fails to match the reference pattern associated with the crimp type.

[0218] E79. The method of E78, wherein the silhouette image fails to match the reference pattern associated with the crimp type when the plurality of zones fail to be identified in the silhouette image.

[0219] E80. The method of any of E53-E79, wherein the defect is one or more of a missing stopper, a missing cap, or a missing flip-off.Electronically Filed May 6, 2026

[0220] E81. The method of any of E53-E80, wherein the silhouette image is a first silhouette image, the camera is a first camera, and the backlight is a first backlight and further comprising: generating a second silhouette image of the vial located between a second camera and a second backlight at the defect detection station of the assembly line system; identifying, via one or more processors, the plurality of zones within the second silhouette image based on the reference pattern associated with the crimp type; analyzing, via the one or more processors, the second silhouette image to measure the plurality of evaluation metrics for the seal crimp of the vial with respect to the plurality of zones; and determining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that at least one of the plurality evaluation metrics measured from the second silhouette image exhibits the corresponding alert condition.

[0221] E82. The method of E81, wherein the first camera and the second camera are mounted to a single bracket.

[0222] E83. The method of E81 or E82, wherein the first camera and the second camera are oriented relative to each other at an angle in a range of 65-95 degrees.

[0223] E84. The method of any of E81-E83, further comprising: averaging one of the plurality evaluation metrics measured from the first silhouette image with a corresponding one of the plurality evaluation metrics measured from the second silhouette image to generate an average metric; and determining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that the average metric exhibits the corresponding alert condition.

[0224] E85. The method of any of E81-E84, further comprising: combining one of the plurality evaluation metrics measured from the first silhouette image with a corresponding one of the plurality evaluation metrics measured from the second silhouette image to generate a composite metric; and determining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that the composite metric exhibits the corresponding alert condition.

[0225] E86. The method of E85, wherein a multi-metric alert condition is defined by defining, via the one or more processors, respective composite alert metric threshold values for the plurality of evaluation metrics, and the multi-metric alert condition is exhibited when one or more of the measurements of the composite metrics are outside the respective composite alert metric threshold value.

[0226] E87. The method of any of E81-E86, wherein: the plurality of zones include: a right top zone referenced to a top right portion of a flip-off portion of the seal crimp; a leftElectronically Filed May 6, 2026 top zone referenced to a top left portion of the flip-off portion of the seal crimp; a right side zone referenced to a right side portion of the seal crimp; a left side zone referenced to a left side portion of the seal crimp; and a right bottom zone referenced to a bottom right portion of the seal crimp; and the plurality of evaluation metrics include: a right side angle between a right edge location in the right side zone and a first right bottom edge location in the right bottom zone; a left side angle between a left edge location in the left side zone and a first left bottom edge location in the left bottom zone; a right side distance between a top edge location of the right top zone and a second bottom edge location in the right bottom zone; a left side distance between a top edge location of the left top zone and a second bottom edge location in the left bottom zone; a bottom right area within the right bottom zone; and a bottom left area within the left bottom zone.

[0227] E88. The method of E87, further comprising: determining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that (i) more than one of the right side angle, the left side angle, the bottom right area, and the bottom left area as measured in either the first silhouette image or the second silhouette image is outside a secondary threshold and (ii) a sum of the left side distance and right side distance measured from the first silhouette image and the left side distance and right side distance measured from the second silhouette image is outside a composite distance threshold.

[0228] E89. The method of any of E53-E88, wherein rejecting the vial includes directing the vial to a defective vial location of the assembly line system.

[0229] E90. A defect detection system comprising: at least one camera; at least one backlight arranged telecentrically with the at least one camera; a memory storing a reference pattern associated with a vial crimp type; and a control unit having one or more processors operatively connected with the at least one camera and the memory and one or more non-transitory, computer-readable media storing instructions that, when executed by the one or more processors, cause the control unit to: receive a silhouette image of a vial generated by the at least one camera while the vial is located between the at least one camera and the at least one backlight, the vial having the vial crimp type; identify a plurality of zones within the silhouette image, wherein the plurality of zones are defined based on a reference pattern associated with the crimp type; analyze the silhouette image to measure a plurality of evaluation metrics for a seal crimp of the vial, wherein the measurements are based on image data within the identified plurality of zones; determine that the seal crimp of the vial includes a defect by determining that at least one of the plurality of evaluation metrics exhibits aElectronically Filed May 6, 2026 corresponding alert condition; and generate an alert signal indicating that the vial is defective based on whether the seal crimp of the vial includes the defect.

[0230] E91. The defect detection system of E90, wherein the pattern associated with the crimp type is defined by: generating a representative silhouette image of a crimp defect free vial having the crimp type; defining a size and relative location of the plurality of zones within the representative silhouette image; and saving the defined zones as the reference pattern associated with the crimp type.

[0231] E92. The defect detection system of E90 or E91, wherein the reference pattern associated with the crimp type is defined by: defining the plurality of evaluation metrics with respect to the plurality of zones; and associating the defined plurality of evaluation metrics with the reference pattern.

[0232] E93. The defect detection system of any of E90-E92, wherein the alert condition is defined by defining respective primary threshold values for each of the evaluation metrics, wherein the alert conditions include one or more of the plurality of evaluation metrics outside the respective primary threshold values.

[0233] E94. The defect detection system of any of E90-E93, wherein the alert conditions are defined by: generating a representative set of defective values for each evaluation metric and a representative set of non-defective values for each evaluation metric; and defining respective primary threshold values that delineate the representative sets of defective values and the representative sets of non-defective values for each of the evaluation metrics, wherein the alert conditions include one or more of the plurality of evaluation metrics outside the respective primary threshold values.

[0234] E95. The defect detection system of E94, wherein generating the representative set of defective values for each evaluation metric and the representative set of non-defective values for each evaluation metric includes: generating a first set of testing silhouette images for defect free vials having the crimp type; generating a second set of testing silhouette images for defective crimp vials having the crimp type; identifying the plurality of zones within the first set of silhouette images and the second set of silhouette images based on the reference pattern associated with the crimp type; and measuring, within the first set of silhouette images and the second set of silhouette images, the plurality of evaluation metrics for seal crimps of the defect free vials and the defective crimp vials with respect to the plurality of zones to generate the representative set of defective values and the representative set of non-defective values.Electronically Filed May 6, 2026

[0235] E96. The defect detection system of any of E90-E95, wherein the alert condition includes a multi-metric alert condition.

[0236] E97. The defect detection system of E96, wherein the multi -metric alert condition includes a Boolean rule.

[0237] E98. The defect detection system of E97, wherein the Boolean rule is determined using a greedy disjunctive normal form algorithm.

[0238] E99. The defect detection system of E97 or E98, wherein the Boolean rule is determined using an artificial intelligence algorithm.

[0239] El 00. The defect detection system of any of E96-E99, wherein the multi-metric alert condition is defined by defining respective secondary threshold values for the plurality of evaluation metrics, and the multi-metric alert condition is exhibited when two or more of the measurements of the plurality of evaluation metrics are outside the respective secondary threshold values.

[0240] E101. The defect detection system of any of E94-E100, wherein: the alert condition includes a multi-metric alert condition for evaluating the plurality of evaluation metrics; the multi-metric alert condition is defined by defining respective secondary threshold values for the plurality of evaluation metrics that also delineate the representative sets of defective values and the representative sets of non-defective values for each of the evaluation metrics; and the multi-metric alert condition is exhibited when two of the measurements of the plurality of evaluation metrics are outside the respective secondary threshold values.

[0241] El 02. The defect detection system of any of E90-E101, wherein the at least one camera comprises a telecentric lens.

[0242] El 03. The defect detection system of any of E90-E102, wherein the plurality of zones include at least one of: a top zone referenced to a top portion of a flip-off portion of the seal crimp; a side zone referenced to a side portion of the seal crimp; and a bottom zone referenced to a bottom portion of the seal crimp.

[0243] E104. The defect detection system of E103, wherein the plurality of evaluation metrics include an angle between a first edge location within the side zone and a second edge location within the bottom zone.

[0244] E105. The defect detection system of E103 or E104, wherein the plurality of evaluation metrics include a plurality of angles, each angle defined between a first edge location within a side zone and one of a plurality of edge locations within the bottom zone.

[0245] E106. The defect detection system of any of E103-E105, wherein the plurality of evaluation metrics include an angle between a first edge location within the side zone and aElectronically Filed May 6, 2026 second edge location within the bottom zone and the angle is measured by: scanning the side zone for changes in pixel values that identify the first edge; scanning the bottom zone for changes in pixel values that identify the second edge; and measuring the angle between the first edge and the second edge.

[0246] E107. The defect detection system of any of E103-E106, wherein the plurality of evaluation metrics include an area within the bottom zone.

[0247] E108. The defect detection system of any of E103-E107, wherein the plurality of evaluation metrics include an area within the bottom zone and the area within the bottom zone is measured by: measuring the area of pixels within a silhouette of the vial located within the bottom zone.

[0248] E109. The defect detection system of any of E103-E108, wherein the plurality of evaluation metrics include a radius of an arc fitted to a first edge location within the side zone and a second edge location within the bottom zone.

[0249] E110. The defect detection system of any of E103-E109, wherein the plurality of evaluation metrics include a distance measured between a crimp edge location and an intersection of a first edge location within the side zone and a second edge location within the bottom zone.

[0250] El 11. The defect detection system of any of E103-E110, wherein the plurality of evaluation metrics include a plurality of intersection distance metrics at a plurality of crimp edge locations, each intersection distance metric measured between one of the plurality of crimp edge locations and an intersection of a first edge location within the side zone and a second edge location within the bottom zone.

[0251] El 12. The defect detection system of any of E103-E111, wherein: the side zone is a first side zone and the bottom zone is a first bottom zone; the plurality of zones include a second side zone referenced to another side portion of the seal crimp and a second bottom zone referenced to another bottom portion of the seal crimp; the plurality of evaluation metrics include a difference between a first angle and a second angle; the first angle is between a first edge location in the first side zone and a second edge location in the first bottom zone; and the second angle is between a third edge location in the second side zone and a fourth edge location in the second bottom zone.

[0252] El 13. The defect detection system of El 12, wherein the difference is determined by: scanning the first side zone for changes in pixel values that identify the first edge; scanning the first bottom zone for changes in pixel values that identify the second edge; scanning the second side zone for changes in pixel values that identify the third edge;Electronically Filed May 6, 2026 scanning the second bottom zone for changes in pixel values that identify the fourth edge; measuring the first angle between the first edge and the second edge and the second angle between the third edge and the fourth edge; and subtracting the first angle from the second angle to determine the difference.

[0253] El 14. The defect detection system of any of E90-E113, wherein the instructions, when executed by the one or more processors, cause the control unit to: determine that the seal crimp of the vial includes the defect by determining that the silhouette image fails to match the reference pattern associated with the crimp type.

[0254] El 15. The defect detection system of El 14, wherein the silhouette image fails to match the reference pattern associated with the crimp type when the one or more processors fail to identify the plurality of zones in the silhouette image.

[0255] El 16. The defect detection system of any of E90-E115, wherein the defect is one or more of a missing stopper, a missing cap, or a missing flip-off.

[0256] El 17. The defect detection system of any of E90-E116, wherein the silhouette image is a first silhouette image, the at least one camera is a first camera, and the at least one backlight is a first backlight and further comprising: a second camera; and a second backlight arranged telecentrically with the second camera, wherein the instructions, when executed by the one or more processors, cause the control unit to: receive a second silhouette image of the vial located between the second camera and the second backlight; identify the plurality of zones within the second silhouette image based on the reference pattern associated with the crimp type; analyze the second silhouette image to measure the plurality of evaluation metrics for the seal crimp of the vial with respect to the plurality of zones; and determine that the seal crimp of the vial includes the defect by determining that at least one of the plurality evaluation metrics measured from the second silhouette image exhibits the corresponding alert condition.

[0257] El 18. The defect detection system of El 17, wherein the first camera and the second camera are mounted to a single bracket.

[0258] El 19. The defect detection system of El 17 or El 18, wherein the first camera and the second camera are oriented relative to each other at an angle in a range of 65-95 degrees.

[0259] E120. The defect detection system of any of El 17-E119, wherein the instructions, when executed by the one or more processors, cause the control unit to: average one of the plurality evaluation metrics measured from the first silhouette image with a corresponding one of the plurality evaluation metrics measured from the second silhouette image to generateElectronically Filed May 6, 2026 an average metric; and determine that the seal crimp of the vial includes the defect by determining that the average metric exhibits the corresponding alert condition.

[0260] E121. The defect detection system of any of El 17-E120, wherein the instructions, when executed by the one or more processors, cause the control unit to: combine one of the plurality evaluation metrics measured from the first silhouette image with a corresponding one of the plurality evaluation metrics measured from the second silhouette image to generate a composite metric; and determine that the seal crimp of the vial includes the defect by determining that the composite metric exhibits the corresponding alert condition.

[0261] E122. The defect detection system of E121, wherein a multi-metric alert condition is defined by defining respective composite alert metric threshold values for the plurality of evaluation metrics, and the multi-metric alert condition is exhibited when one or more of the measurements of the composite metrics are outside the respective composite alert metric threshold value.

[0262] E123. The defect detection system of any of El 17-E122, wherein: the plurality of zones include: a right top zone referenced to a top right portion of a flip-off portion of the seal crimp; a left top zone referenced to a top left portion of the flip-off portion of the seal crimp; a right side zone referenced to a right side portion of the seal crimp; a left side zone referenced to a left side portion of the seal crimp; and a right bottom zone referenced to a bottom right portion of the seal crimp; and the plurality of evaluation metrics include: a right side angle between a right edge location in the right side zone and a first right bottom edge location in the right bottom zone; a left side angle between a left edge location in the left side zone and a first left bottom edge location in the left bottom zone; a right side distance between a top edge location of the right top zone and a second bottom edge location in the right bottom zone; a left side distance between a top edge location of the left top zone and a second bottom edge location in the left bottom zone; a bottom right area within the right bottom zone; and a bottom left area within the left bottom zone.

[0263] E124. The defect detection system of E123, wherein the instructions, when executed by the one or more processors, cause the control unit to: determine that the seal crimp of the vial includes the defect by determining that (i) more than one of the right side angle, the left side angle, the bottom right area, and the bottom left area as measured in either the first silhouette image or the second silhouette image is outside a secondary threshold and (ii) a sum of the left side distance and right side distance measured from the first silhouette image and the left side distance and right side distance measured from the second silhouette image is outside a composite distance threshold.Electronically Filed May 6, 2026

[0264] E125. An assembly line system comprising: the defect detection system of any of E90-E124 deployed at a defect detection station; and an assembly line controller configured to reject the vial by receiving the alert signal from the defect detection system.

[0265] E126. The assembly line system of E125, wherein to reject the vial the assembly line controller directs the vial to a defective vial location of the assembly line system.

[0266] E127. A method comprising: generating a silhouette image of a vial located between a camera and a backlight at a defect detection station of an assembly line system, the vial having an associated crimp type; determining, via one or more processors, whether the silhouette image matches a reference pattern associated with the crimp type, the reference pattern defining a plurality of zones to be located within the silhouette image; determining, via the one or more processors, whether the seal crimp of the vial includes a defect by determining whether the silhouette image fails to match the reference pattern associated with the crimp type; and rejecting or accepting the vial based on whether the seal crimp of the vial includes the defect.

[0267] E128. The method of E127, wherein the silhouette image fails to match the reference pattern associated with the crimp type when the plurality of zones fail to be identified in the silhouette image.

[0268] E129. The method of E127 or E128, further comprising: identifying, via the one or more processors, the plurality of zones within the silhouette image; analyzing, via the one or more processors, the silhouette image to measure a plurality of evaluation metrics for the seal crimp of the vial, wherein the measurements are based on image data within the identified plurality of zones; and determining, via the one or more processors, whether the seal crimp of the vial includes the defect by determining whether at least one of the plurality of evaluation metrics exhibits a corresponding alert condition.

[0269] E130. A defect detection system comprising: at least one camera; at least one backlight arranged telecentrically with the at least one camera; a memory storing a reference pattern associated with a vial crimp type; and a control unit having one or more processors operatively connected with the at least one camera and the memory and one or more non-transitory, computer-readable media storing instructions that, when executed by the one or more processors, cause the control unit to: receive a silhouette image of a vial generated by the at least one camera while the vial is located between the at least one camera and the at least one backlight, the vial having the vial crimp type; determine whether the silhouette image matches a reference pattern associated with the crimp type, the reference pattern defining a plurality of zones to be located within the silhouette image; determine whether theElectronically Filed May 6, 2026 seal crimp of the vial includes a defect by determining whether the silhouette image fails to match the reference pattern associated with the crimp type; and generate an alert signal indicating that the vial is defective based on whether the seal crimp of the vial includes the defect.

[0270] E131. The defect detection system of E130, wherein the silhouette image fails to match the reference pattern associated with the crimp type when the plurality of zones fail to be identified in the silhouette image.

[0271] E132. The defect detection system of E130 or E131, wherein the instructions, when executed by the one or more processors, cause the control unit to: identify the plurality of zones within the silhouette image; analyze the silhouette image to measure a plurality of evaluation metrics for the seal crimp of the vial, wherein the measurements are based on image data within the identified plurality of zones; and determine whether the seal crimp of the vial includes the defect by determining whether at least one of the plurality of evaluation metrics exhibits a corresponding alert condition.

[0272] Although the devices, systems, assemblies, components, subsystems and methods have been described in terms of exemplary embodiments, they are not limited thereto. The detailed description is to be construed as exemplary only and does not describe every possible embodiment of the present disclosure. Numerous alternative embodiments could be implemented, using either current technology or technology developed after the filing date of this patent that would still fall within the scope of the claims defining the invention(s) disclosed herein.

[0273] Those skilled in the art will recognize that a wide variety of modifications, alterations, and combinations can be made with respect to the above-described embodiments without departing from the spirit and scope of the invention(s) disclosed herein, and that such modifications, alterations, and combinations are to be viewed as being within the ambit of the inventive concept(s).

Claims

Electronically Filed May 6, 2026CLAIMS1. A method comprising:generating a silhouette image of a vial located between a camera and a backlight at a defect detection station of an assembly line system, the vial having an associated crimp type;identifying, via one or more processors, a plurality of zones within the silhouette image, wherein the plurality of zones are defined based on a reference pattern associated with the crimp type;analyzing, via the one or more processors, the silhouette image to measure a plurality of evaluation metrics for a seal crimp of the vial, wherein the measurements are based on image data within the identified plurality of zones;determining, via the one or more processors, whether the seal crimp of the vial includes a defect by determining whether at least one of the plurality of evaluation metrics exhibits a corresponding alert condition; andrejecting or accepting the vial based on whether the seal crimp of the vial includes the defect.

2. The method of claim 1, further comprising defining the reference pattern associated with the crimp type by:generating a representative silhouette image of a crimp defect free vial having the crimp type;defining, via the one or more processors, a size and relative location of the plurality of zones within the representative silhouette image; andsaving, via the one or more processors, the defined zones as the reference pattern associated with the crimp type.

3. The method of any one of claims 1-2, further comprising defining the alert condition by: defining, via the one or more processors, respective primary threshold values for each ofElectronically Filed May 6, 2026 the evaluation metrics, wherein the alert conditions include one or more of the plurality of evaluation metrics outside the respective primary threshold values.

4. The method of any one of claims 1-3, wherein the alert condition includes a multimetric alert condition.

5. The method of claim 4, wherein the multi -metric alert condition includes a Boolean rule.

6. The method of claim 5, wherein the Boolean rule is determined using a greedy disjunctive normal form algorithm7. The method of any one of claims 5-6, wherein the Boolean rule is determined using an artificial intelligence algorithm.

8. The method of any one of claims 4-7, wherein the multi-metric alert condition is defined by defining, via the one or more processors, respective secondary threshold values for the plurality of evaluation metrics, and the multi-metric alert condition is exhibited when two or more of the measurements of the plurality of evaluation metrics are outside the respective secondary threshold values.

9. The method of any one of claims 1-8, wherein:the alert condition includes a multi-metric alert condition for evaluating the plurality of evaluation metrics;the multi-metric alert condition is defined by defining, via the one or more processors, respective secondary threshold values for the plurality of evaluation metrics that also delineate representative sets of defective values and representative sets of non-defective values for each of the evaluation metrics; andElectronically Filed May 6, 2026 the multi-metric alert condition is exhibited when two of the measurements of the plurality of evaluation metrics are outside the respective secondary threshold values.

10. The method of any one of claims 1-9, wherein the camera comprises a telecentric lens.

11. The method of any one of claims 1-10, wherein the plurality of zones include at least one of:a top zone referenced to a top portion of a flip-off portion of the seal crimp;a side zone referenced to a side portion of the seal crimp; anda bottom zone referenced to a bottom portion of the seal crimp.

12. The method of claim 11, wherein the plurality of evaluation metrics include an angle between a first edge location within the side zone and a second edge location within the bottom zone.

13. The method of any one of claims 11-12, wherein the plurality of evaluation metrics include a plurality of angles, each angle defined between a first edge location within a side zone and one of a plurality of edge locations within the bottom zone.

14. The method of any one of claims 11-13, wherein the plurality of evaluation metrics include an area within the bottom zone.

15. The method of any one of claims 11-14, wherein the plurality of evaluation metrics include a radius of an arc fitted to a first edge location within the side zone and a second edge location within the bottom zone.Electronically Filed May 6, 2026 16. The method of any one of claims 11-15, wherein the plurality of evaluation metrics include a distance measured between a crimp edge location and an intersection of a first edge location within the side zone and a second edge location within the bottom zone.

17. The method of any one of claims 11-16, wherein the plurality of evaluation metrics include a plurality of intersection distance metrics at a plurality of crimp edge locations, each intersection distance metric measured between a one of the plurality of crimp edge locations and an intersection of a first edge location within the side zone and a second edge location within the bottom zone.

18. The method of any one of claims 1-17, further comprising:determining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that the silhouette image fails to match the reference pattern associated with the crimp type.

19. The method of claim 18, wherein the silhouette image fails to match the reference pattern associated with the crimp type when the plurality of zones fail to be identified in the silhouette image.

20. The method of any one of claims 1-19, wherein the defect is one or more of a missing stopper, a missing cap, or a missing flip-off.

21. The method of any one of claims 1-20, wherein the silhouette image is a first silhouette image, the camera is a first camera, and the backlight is a first backlight and further comprising:generating a second silhouette image of the vial located between a second camera and a second backlight at the defect detection station of the assembly line system;identifying, via one or more processors, the plurality of zones within the second silhouette image based on the reference pattern associated with the crimp type;Electronically Filed May 6, 2026 analyzing, via the one or more processors, the second silhouette image to measure the plurality of evaluation metrics for the seal crimp of the vial with respect to the plurality of zones; anddetermining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that at least one of the plurality evaluation metrics measured from the second silhouette image exhibits the corresponding alert condition.

22. The method of claim 21, wherein the first camera and the second camera are oriented relative to each other at an angle in a range of 65-95 degrees.

23. The method of any one of claims 21-22, further comprising:averaging one of the plurality evaluation metrics measured from the first silhouette image with a corresponding one of the plurality evaluation metrics measured from the second silhouette image to generate an average metric; anddetermining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that the average metric exhibits the corresponding alert condition.

24. The method of any one of claims 21-23, further comprising:combining one of the plurality evaluation metrics measured from the first silhouette image with a corresponding one of the plurality evaluation metrics measured from the second silhouette image to generate a composite metric; anddetermining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that the composite metric exhibits the corresponding alert condition.

25. The method of claim 24, wherein a multi-metric alert condition is defined by defining, via the one or more processors, respective composite alert metric threshold values for the plurality of evaluation metrics, and the multi-metric alert condition is exhibited when one or more of the measurements of the composite metrics are outside the respective composite alert metric threshold valueElectronically Filed May 6, 202626. A defect detection system comprising:at least one camera;at least one backlight arranged telecentrically with the at least one camera;a memory storing a reference pattern associated with a vial crimp type; and a control unit having one or more processors operatively connected with the at least one camera and the memory and one or more non-transitory, computer-readable media storing instructions that, when executed by the one or more processors, cause the control unit to:receive a silhouette image of a vial generated by the at least one camera while the vial is located between the at least one camera and the at least one backlight, the vial having the vial crimp type;identify a plurality of zones within the silhouette image, wherein the plurality of zones are defined based on a reference pattern associated with the crimp type;analyze the silhouette image to measure a plurality of evaluation metrics for a seal crimp of the vial, wherein the measurements are based on image data within the identified plurality of zones;determine that the seal crimp of the vial includes a defect by determining that at least one of the plurality of evaluation metrics exhibits a corresponding alert condition; and generate an alert signal indicating that the vial is defective based on whether the seal crimp of the vial includes the defect.

27. The defect detection system of claim 26, wherein the pattern associated with the crimp type is defined by:generating a representative silhouette image of a crimp defect free vial having the crimp type;defining a size and relative location of the plurality of zones within the representative silhouette image; andElectronically Filed May 6, 2026 saving the defined zones as the reference pattern associated with the crimp type.

28. The defect detection system of any one of claims 26-27, wherein the alert condition is defined by defining, via the one or more processors, respective primary threshold values for each of the evaluation metrics, wherein the alert conditions include one or more of the plurality of evaluation metrics outside the respective primary threshold values.

29. The defect detection system of any one of claims 26-28, wherein the alert condition includes a multi-metric alert condition.

30. The defect detection system of claim 29, wherein the multi-metric alert condition includes a Boolean rule.

31. The defect detection system of claim 30, wherein the Boolean rule isdetermined using a greedy disjunctive normal form algorithm32. The defect detection system of any one of claims 30-31, wherein the Boolean rule is determined using an artificial intelligence algorithm.

33. The defect detection system of any one of claims 29-32, wherein defining the multimetric alert condition includes defining, via the one or more processors, respective secondary threshold values for the plurality of evaluation metrics, and the multi-metric alert condition is exhibited when two or more of the measurements of the plurality of evaluation metrics are outside the respective secondary threshold values.

34. The defect detection system of any one of claims 26-33, wherein:the alert condition includes a multi-metric alert condition for evaluating the plurality of evaluation metrics;defining the multi-metric alert condition includes defining, via the one or more processors, respective secondary threshold values for the plurality of evaluation metrics thatElectronically Filed May 6, 2026 also delineate representative sets of defective values and representative sets of non-defective values for each of the evaluation metrics; andthe multi-metric alert condition is exhibited when two of the measurements of the plurality of evaluation metrics are outside the respective secondary threshold values.

35. The defect detection system of any one of claims 26-34, wherein the at least one camera comprises a telecentric lens.

36. The defect detection system of any one of claims 26-35, wherein the plurality of zones include at least one of:a top zone referenced to a top portion of a flip-off portion of the seal crimp;a side zone referenced to a side portion of the seal crimp; anda bottom zone referenced to a bottom portion of the seal crimp.

37. The defect detection system of claim 36, wherein the plurality of evaluation metrics include an angle between a first edge location within the side zone and a second edge location within the bottom zone.

38. The defect detection system of any one of claims 36-37, wherein the plurality of evaluation metrics include a plurality of angles, each angle defined between a first edge location within a side zone and one of a plurality of edge locations within the bottom zone.

39. The defect detection system of any one of claims 36-38, wherein the plurality of evaluation metrics include an area within the bottom zone.Electronically Filed May 6, 2026 40. The defect detection system of any one of claims 36-39, wherein the plurality of evaluation metrics include a radius of an arc fitted to a first edge location within the side zone and a second edge location within the bottom zone.

41. The defect detection system of any one of claims 36-40, wherein the plurality of evaluation metrics include a distance measured between a crimp edge location and an intersection of a first edge location within the side zone and a second edge location within the bottom zone.

42. The defect detection system of any one of claims 36-41, wherein the plurality of evaluation metrics include a plurality of intersection distance metrics at a plurality of crimp edge locations, each intersection distance metric measured between one of the plurality of crimp edge locations and an intersection of a first edge location within the side zone and a second edge location within the bottom zone along a defined direction.

43. The defect detection system of any one of claims 26-42, wherein the instructions, when executed by the one or more processors, cause the control unit to:determine that the seal crimp of the vial includes the defect by determining that the silhouette image fails to match the reference pattern associated with the crimp type.

44. The defect detection system of claim 43, wherein the silhouette image fails to match the reference pattern associated with the crimp type when the one or more processors fail to identify the plurality of zones in the silhouette image.

45. The defect detection system of any one of claims 26-44, wherein the defect is one or more of a missing stopper, a missing cap, or a missing flip-off.Electronically Filed May 6, 2026 46. The defect detection system of any one of claims 26-45, wherein the silhouette image is a first silhouette image, the at least one camera is a first camera, and the at least one backlight is a first backlight and further comprising:a second camera; anda second backlight arranged telecentrically with the second camera,wherein, the instructions, when executed by the one or more processors, cause the control unit to:receive a second silhouette image of the vial located between the second camera and the second backlight;identify the plurality of zones within the second silhouette image based on the reference pattern associated with the crimp type;analyze the second silhouette image to measure the plurality of evaluation metrics for the seal crimp of the vial with respect to the plurality of zones; anddetermine that the seal crimp of the vial includes the defect by determining that at least one of the plurality evaluation metrics measured from the second silhouette image exhibits the corresponding alert condition.

47. The defect detection system of claim 46, wherein the first camera and the second camera are mounted to a single bracket.

48. The defect detection system of any one of claims 46-47, wherein the first camera and the second camera are oriented relative to each other at an angle in a range of 65-95 degrees.

49. The defect detection system of any one of claims 46-48, wherein the instructions, when executed by the one or more processors, cause the control unit to:average one of the plurality evaluation metrics measured from the first silhouette image with a corresponding one of the plurality evaluation metrics measured from the second silhouette image to generate an average metric; andElectronically Filed May 6, 2026 determine that the seal crimp of the vial includes the defect by determining that the average metric exhibits the corresponding alert condition.

50. The defect detection system of any one of claims 46-49, further comprising:combining one of the plurality evaluation metrics measured from the first silhouette image with a corresponding one of the plurality evaluation metrics measured from the second silhouette image to generate a composite metric; anddetermining, via the one or more processors, that the seal crimp of the vial includes the defect by determining that the composite metric exhibits the corresponding alert condition.

51. The defect detection system of claim 50, wherein a multi-metric alert condition is defined by defining, via the one or more processors, respective composite alert metric threshold values for the plurality of evaluation metrics, and the multi-metric alert condition is exhibited when one or more of the measurements of the composite metrics are outside the respective composite alert metric threshold value.

52. An assembly line system comprising:the silhouette imaging and defect detection system of any one of claims 26-51 deployed at a defect detection station; andan assembly line controller configured to reject the vial by receiving the alert signal from the defect detection system.

53. The assembly line system of claim 52, wherein to reject the vial assembly line controller directs the vial to a defective vial location of the assembly line system.

54. A method comprising:Electronically Filed May 6, 2026 generating a silhouette image of a vial located between a camera and a backlight at a defect detection station of an assembly line system, the vial having an associated crimp type;determining, via one or more processors, whether the silhouette image matches a reference pattern associated with the crimp type, the reference pattern defining a plurality of zones to be located within the silhouette image;determining, via the one or more processors, whether the seal crimp of the vial includes a defect by determining whether the silhouette image fails to match the reference pattern associated with the crimp type; andrejecting or accepting the vial based on whether the seal crimp of the vial includes the defect.

55. A defect detection system comprising:at least one camera;at least one backlight arranged telecentrically with the at least one camera;a memory storing a reference pattern associated with a vial crimp type; and a control unit having one or more processors operatively connected with the at least one camera and the memory and one or more non-transitory, computer-readable media storing instructions that, when executed by the one or more processors, cause the control unit to:receive a silhouette image of a vial generated by the at least one camera while the vial is located between the at least one camera and the at least one backlight, the vial having the vial crimp type;determine whether the silhouette image matches a reference pattern associated with the crimp type, the reference pattern defining a plurality of zones to be located within the silhouette image;determine whether the seal crimp of the vial includes a defect by determining whether the silhouette image fails to match the reference pattern associated with the crimp type; andElectronically Filed May 6, 2026 generate an alert signal indicating that the vial is defective based on whether the seal crimp of the vial includes the defect.