A silicon lens appearance defect detection device, detection method and detection system

Abstract

This invention relates to the field of optical component appearance defect detection technology, and particularly to a detection device, method, and system for silicon lens appearance defects. The system acquires a reference image and a detection image of the silicon lens wafer through a scanning mechanism. Based on the reference image, a coordinate mapping relationship is established between the arrangement position of the silicon lens units in the wafer map coordinate system and the target position of the motion mechanism. Then, based on this coordinate mapping relationship, the motion mechanism is controlled to drive the scanning mechanism to scan along the scanning path. During the scanning process, the current or subsequent target position is corrected based on the center deviation of the silicon lens units in the detection image. The appearance defect detection result is obtained based on the detection image, thereby improving the stability of the correspondence between the scanning position and the actual position of the silicon lens units, reducing the accumulation of positional deviations in continuous scanning, and making the defect detection results more reliable.

Description

Technical Field

[0001] This invention relates to the field of optical component appearance defect detection technology, and in particular to a detection device, detection method and detection system for appearance defects of silicon lenses. Background Technology

[0002] Silicon lenses, as core micro-optical components in the fields of optical communication modules and optoelectronic co-packaging (CPO), mainly undertake the functions of beam collimation, focusing and optical path coupling, and are key devices to ensure stable transmission of high-speed optical communication signals.

[0003] Silicon lenses are typically mass-produced in wafer form. A large number of silicon lens units are distributed on the same wafer, arranged in rows and columns. Before leaving the factory or entering subsequent processes, it is necessary to inspect each silicon lens unit for appearance defects such as pits, dents, chipping, and scratches.

[0004] In the actual testing process, the testing equipment needs to scan the entire wafer according to the product distribution on the silicon lens wafer, and make the scanning unit align with the silicon lens unit at different positions in sequence to obtain the test image. For wafer-level array products, the test result depends not only on the image recognition capability, but also on whether the target position can correspond to the actual arrangement position of the silicon lens unit during the scanning process.

[0005] In existing silicon lens wafer appearance inspection methods, there is still a problem that it is difficult to accurately map the arrangement of silicon lens units in the wafer map coordinate system to the target position of the motion mechanism and to correct the target position deviation online during the whole wafer scanning process. This directly affects the consistency of the image acquisition position of each silicon lens unit, and thus affects the stability of the appearance defect detection results and the reliability of the subsequent output results. Summary of the Invention

[0006] The main objective of this invention is to provide a detection device, method, and system for detecting defects in the appearance of silicon lenses, aiming to solve the problem that existing detection devices have difficulty accurately mapping the arrangement positions in the wafer map coordinate system to the target positions of the moving mechanism, thus affecting the detection accuracy.

[0007] To achieve the above objectives, the present invention provides a device for detecting appearance defects in silicon lenses, comprising: Bearing mechanism A fixing mechanism is disposed on a support mechanism and is used to fix a silicon lens wafer. A motion mechanism is disposed on a support mechanism, the motion mechanism includes a first motion unit that moves along a first direction and a second motion unit that moves along a second direction, the second motion unit being disposed in the movable part of the first motion unit; A scanning mechanism is connected to the second motion unit. The scanning mechanism includes an illumination unit and a scanning unit, with the scanning unit positioned above the silicon lens wafer and the illumination unit positioned above and / or below the silicon lens wafer.

[0008] Optionally, the fixing mechanism includes a suction cup unit, the upper end face of which is provided with suction holes and a plurality of adsorption tracks. The adsorption tracks are connected to a vacuum pump through the suction holes, and the plurality of adsorption tracks are interconnected to form an adsorption area on the upper end face of the suction cup unit.

[0009] Optionally, the lighting unit includes a first lighting source and a second lighting source, wherein the first lighting source is disposed above the silicon lens wafer, and the second lighting source is disposed below the silicon lens wafer and located inside the suction cup unit.

[0010] Optionally, the first motion unit includes two driving sub-units arranged opposite each other, the two ends of the second motion unit are respectively connected to the movable parts of the two driving sub-units, and the scanning mechanism is connected to the movable part of the second motion unit so that the scanning mechanism performs two-dimensional scanning on the silicon lens wafer.

[0011] To achieve the above objectives, the present invention also provides a method for detecting appearance defects in silicon lenses, comprising the following steps: A reference mark image is acquired at multiple preset reference mark positions on a silicon lens wafer; based on the reference mark image, image positioning information within the reference mark field of view and the physical coordinates of the motion mechanism when acquiring the corresponding reference mark image are obtained; wherein, the image positioning information includes the actual image position of the reference mark and the center position of the lens, and a plurality of silicon lens units are distributed on the silicon lens wafer; Based on the image positioning information and the physical coordinates, an initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system is obtained; and based on the initial correspondence, a two-dimensional coordinate mapping model between the wafer map coordinate system and the motion mechanism coordinate system is constructed. The location of the scanning start area and the product distribution of the silicon lens wafer are obtained. Based on the location of the scanning start area, the product distribution of the silicon lens wafer, and the two-dimensional coordinate mapping model, the scanning path covering the silicon lens wafer is obtained. Based on the scanning path and the two-dimensional coordinate mapping model, the theoretical target position corresponding to each target scanning area is obtained; based on the theoretical target position, the motion mechanism is controlled to drive the scanning unit to move sequentially to each target scanning area along the scanning path, and the corresponding target scanning area image is obtained; Based on the target scanning area image, obtain the product image corresponding to each silicon lens unit, obtain the defect feature parameters based on the product image, and obtain the silicon lens appearance defect detection result based on the defect feature parameters.

[0012] Optionally, after acquiring the corresponding target scan area image, the following steps are included: During the scanning process, at predetermined intervals, a target scanning area containing the central product is selected from each target scanning area as a re-anchoring scanning area; based on the target scanning area image corresponding to the re-anchoring scanning area, the actual image center position of the central product is obtained; based on the actual image center position and the current field-of-view image center position of the scanning unit, the pixel deviation is obtained; based on the pixel deviation, the corresponding platform compensation amount is obtained; based on the platform compensation amount, the target platform coordinates corresponding to the current target scanning area or subsequent target scanning areas are corrected to complete online position calibration.

[0013] Optionally, obtaining image positioning information within the field of view of the reference marker and the physical coordinates of the motion mechanism when obtaining the corresponding reference marker image, based on the reference marker image, includes the following steps: The reference marker image is subjected to reference marker recognition to obtain the actual image position of the reference marker; The silicon lens unit in the reference mark image is centrally located to obtain the center position of the lens within the field of view of the reference mark; When the scanning unit acquires the corresponding reference mark image, the physical coordinates of the motion mechanism in the first and second directions are obtained; The actual image position of the reference mark, the center position of the lens, and the physical coordinates of the motion mechanism are correlated to form reference alignment data.

[0014] Optionally, the step of obtaining the initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system based on the image positioning information and the physical coordinates, and constructing a two-dimensional coordinate mapping model between the wafer map coordinate system and the motion mechanism coordinate system based on the initial correspondence, includes the following steps: Based on multiple reference alignment data, obtain the row and column positions of multiple reference points in the wafer map coordinate system and their physical coordinates in the motion mechanism coordinate system; Based on the row and column positions of multiple reference points in the wafer map coordinate system and their physical coordinates in the motion mechanism coordinate system, the initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system is obtained. Based on the initial correspondence, determine the reference row and column positions of the reference reference mark or reference silicon lens unit in the wafer map coordinate system and the reference physical coordinates in the motion mechanism coordinate system; Based on the reference alignment data corresponding to multiple reference points, obtain the row stepping relationship when the wafer map coordinate system increases along the row direction and the column stepping relationship when it increases along the column direction. Based on the reference row and column positions, reference physical coordinates, row stepping relationships, and column stepping relationships, a two-dimensional coordinate mapping model between the wafer map coordinate system and the motion mechanism coordinate system is constructed.

[0015] Optionally, obtaining the location of the scanning start area and the product distribution of the silicon lens wafer, and obtaining the scanning path covering the silicon lens wafer based on the location of the scanning start area, the product distribution of the silicon lens wafer, and the two-dimensional coordinate mapping model, includes the following steps: Based on the product distribution of the silicon lens wafer, obtain the row and column distribution of the silicon lens unit to be tested in the wafer map coordinate system; Determine the starting target scanning area of ​​the scanning path based on the location of the scanning start area; Based on the row and column distribution of the silicon lens unit to be tested in the wafer map coordinate system, adjacent multi-row and multi-column silicon lens units to be tested are divided into the same target scanning area. Based on the two-dimensional coordinate mapping model, the theoretical target positions corresponding to each target scanning area are obtained; The target scanning areas are arranged in a row-by-row manner to obtain the scanning path covering the silicon lens wafer.

[0016] Optionally, the step of controlling the motion mechanism to move the scanning unit sequentially along the scanning path to each target scanning area according to the theoretical target position, and acquiring the corresponding target scanning area image, includes the following steps: Based on the theoretical target position corresponding to each target scanning area, motion control commands are generated; Based on the lighting requirements corresponding to each target scanning area, generate lighting control commands; According to the motion control command, the motion mechanism is controlled to drive the scanning unit to move sequentially to each target scanning area along the scanning path; According to the lighting control command, the lighting unit is controlled to illuminate the target scanning area currently corresponding to the scanning unit; The control scanning unit acquires the corresponding target scanning area image.

[0017] Optionally, obtaining the corresponding platform compensation amount based on the pixel deviation includes the following steps: Obtain the correspondence between pre-calibrated image coordinates and motion mechanism coordinates; Based on the pixel deviation and the correspondence between the image coordinates and the motion mechanism coordinates, the position offset of the pixel deviation in the motion mechanism coordinate system is obtained; The position offset is used as the platform compensation amount.

[0018] Optionally, the step of obtaining product images corresponding to each silicon lens unit based on the target scanning area image, obtaining defect feature parameters based on the product images, and obtaining silicon lens appearance defect detection results based on the defect feature parameters includes the following steps: Based on the product distribution of the silicon lens wafer, determine the silicon lens units contained in the image of each target scanning area; Based on the position of the silicon lens unit in the image of each target scanning area, obtain the product image of the corresponding silicon lens unit; Associate the product image with the row and column positions of the corresponding silicon lens unit in the wafer map coordinate system; Defect regions are extracted from the product image to obtain the defect regions; Based on the defect region, defect feature parameters are obtained, including area features, aspect ratio features, and grayscale contrast features. Based on the defect feature parameters, the appearance defect type and the corresponding defect size are obtained; wherein, the appearance defect type includes at least one of pitting, dents, chipping, and scratches; Based on the type and size of the appearance defect, obtain the good product or defective product judgment result of the corresponding silicon lens unit; Based on the good or bad product determination results of each silicon lens unit, obtain the appearance defect detection results of the silicon lens.

[0019] To achieve the above objectives, the present invention also provides a detection system for appearance defects in silicon lenses, the system comprising: The positioning information acquisition module is used to acquire reference mark images of multiple preset reference mark positions on the silicon lens wafer; based on the reference mark images, it acquires image positioning information within the reference mark field of view and the physical coordinates of the motion mechanism when acquiring the corresponding reference mark image; wherein, the image positioning information includes the actual image position of the reference mark and the center position of the lens, and the silicon lens wafer is distributed with a plurality of silicon lens units; The coordinate mapping model construction module is used to obtain the initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system based on the image positioning information and the physical coordinates; and to construct a two-dimensional coordinate mapping model between the wafer map coordinate system and the motion mechanism coordinate system based on the initial correspondence. The scan path acquisition module is used to acquire the position of the scan start area and the product distribution of the silicon lens wafer. Based on the position of the scan start area, the product distribution of the silicon lens wafer, and the two-dimensional coordinate mapping model, the scan path covering the silicon lens wafer is acquired. The product image acquisition module is used to obtain the theoretical target position corresponding to each target scanning area according to the scanning path and the two-dimensional coordinate mapping model; according to the theoretical target position, it controls the motion mechanism to drive the scanning unit to move sequentially to each target scanning area along the scanning path and acquire the corresponding target scanning area image. The detection result acquisition module is used to acquire product images corresponding to each silicon lens unit based on the target scanning area image, acquire defect feature parameters based on the product images, and acquire silicon lens appearance defect detection results based on the defect feature parameters.

[0020] The beneficial effects that this invention can achieve are as follows: This invention solves the problem of silicon lens wafers by setting a fixing mechanism on a support structure for fixing silicon lens wafers and configuring a motion mechanism that can drive the scanning mechanism to move along a first direction and a second direction. This allows the scanning mechanism to acquire a reference image and a detection image of the silicon lens wafer. Furthermore, based on the reference image, a coordinate mapping relationship is established between the arrangement position of the silicon lens units in the wafer map coordinate system and the target position of the motion mechanism. Then, based on this coordinate mapping relationship, the motion mechanism is controlled to drive the scanning mechanism to scan the silicon lens wafer along the scanning path. During the scanning process, the current target position or the subsequent target position is corrected based on the center deviation of the silicon lens units in the detection image. In wafer appearance inspection, there are problems such as difficulty in accurately converting the wafer map position into the target position of the equipment movement, easy accumulation of target position deviation during whole-wafer scanning, and unstable correspondence between the inspection image and the actual silicon lens unit position. Therefore, the scanning mechanism continuously completes the inspection image acquisition according to the actual arrangement relationship of the silicon lens units, and corrects the target position in a timely manner based on image feedback during the scanning process. This reduces the impact of wafer placement deviation and motion positioning deviation on the image acquisition position, and ensures that the inspection image of each silicon lens unit maintains a more reliable correspondence with its wafer map position. This improves the stability, location traceability, and subsequent judgment reliability of the whole-wafer silicon lens wafer appearance defect inspection results. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the detection device in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the internal structure of the detection device in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the internal structure of the fixing mechanism in Embodiment 1 of the present invention; Figure 4 This is a partial structural diagram of the fixing mechanism in Embodiment 1 of the present invention; Figure 5 This is a schematic diagram of the structure of the first motion unit in Embodiment 1 of the present invention; Figure 6This is a schematic diagram of the scanning unit in Embodiment 1 of the present invention; Figure 7 This is a flowchart of the detection method in Embodiment 2 of the present invention; Figure 8 This is a structural block diagram of the detection system in Embodiment 3 of the present invention.

[0022] Figure label: 1-Bearing mechanism, 2-Fixing mechanism, 3-Motion mechanism, 4-Scanning mechanism, 5-Silicon lens wafer; 21-Suction cup unit, 22-Suction hole, 23-Suction track; 31-First motion unit, 32-Second motion unit, 33-Limiting unit; 41-Illumination unit, 42-Scanning unit; 331 - First limiter, 332 - Second limiter, 333 - Limiting piece; 411 - First lighting source, 412 - Second lighting source; 421 - Camera subunit, 422 - Lens subunit.

[0023] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0025] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0026] In this invention, unless otherwise explicitly specified and limited, the terms "connection," "fixed," etc., should be interpreted broadly. For example, "fixed" can mean a fixed connection, a detachable connection, or an integral part; it can mean a mechanical connection or an electrical connection; it can mean a direct connection or an indirect connection through an intermediate medium; it can mean the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0027] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the meaning of "and / or" throughout the text includes three parallel solutions; for example, "A and / or B" includes solution A, solution B, or a solution where both A and B are satisfied simultaneously. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0028] Example 1: Please refer to the attached document as well. Figures 1 to 6 This embodiment provides a device for detecting appearance defects in silicon lenses, including: Bearing mechanism 1, Fixing mechanism 2, which is disposed on the bearing mechanism 1, is used to fix the silicon lens wafer 5; The motion mechanism 3 is disposed on the support mechanism 1. The motion mechanism 3 includes a first motion unit 31 that moves along a first direction and a second motion unit 32 that moves along a second direction. The second motion unit 32 is disposed on the movable part of the first motion unit 31. The scanning mechanism 4 is connected to the second motion unit 32. The scanning mechanism 4 includes an illumination unit 41 and a scanning unit 42. The scanning unit 42 is disposed above the silicon lens wafer 5, and the illumination unit 41 is disposed above and / or below the silicon lens wafer 5. The device is used to acquire a reference image and a detection image of the silicon lens wafer 5 through the scanning mechanism 4, and to establish a coordinate mapping relationship between the arrangement position of the silicon lens unit in the wafer map coordinate system and the target position of the motion mechanism 3 based on the reference image. Based on the coordinate mapping relationship, the motion mechanism 3 is controlled to drive the scanning mechanism 4 to scan the silicon lens wafer 5 along the scanning path. During the scanning process, the current target position or the subsequent target position is corrected based on the center deviation of the silicon lens unit in the detection image, and the appearance defect detection result of the silicon lens is obtained based on the detection image.

[0029] It should be noted that the surface of silicon lens wafer 5 has minor appearance defects such as pits, dents, chipping, and scratches, which will directly lead to increased optical path loss and decreased coupling efficiency, seriously affecting the reliability and service life of optical communication products. In the current industry environment, high-magnification microscopes are generally used for manual visual inspection. The manual inspection method is not only inefficient and cannot be adapted to the pace of mass production, but the inspection results also rely heavily on the experience and visual state of the personnel, which is highly subjective.

[0030] Based on the above problems, this embodiment provides a detection device for appearance defects of silicon lenses. In one specific embodiment, the supporting mechanism 1 can be a base plate, which can be made of aluminum alloy plate, stainless steel plate or other metal plate with sufficient rigidity. Its surface can be blackened, electroplated or insulated to reduce the impact of stray reflected light on the imaging quality during the detection process. The supporting mechanism 1 is not only used to provide an installation reference, but also to ensure a stable relative positional relationship between the moving mechanism 3 and the fixed mechanism 2, so that when the scanning mechanism 4 moves under the drive of the moving mechanism 3, it can form a repeatable scanning trajectory relative to the silicon lens wafer 5.

[0031] In one specific embodiment, the fixing mechanism 2 is disposed on the supporting mechanism 1 and is used to fix the silicon lens to be tested. The fixing mechanism 2 can be a suction cup unit 21, preferably a transparent suction cup. The transparent suction cup is fixed to the supporting mechanism 1 by structural connectors such as screws, pressure plates or positioning pins, so that the silicon lens wafer 5 is held in a predetermined testing position during the testing process. The transparent suction cup is preferably made of transparent acrylic material. Its upper surface is used to support the silicon lens wafer 5, and a vacuum suction hole 22 can be provided on the side. The upper surface can be provided with adsorption tracks 23 distributed in a cross or ring. The adsorption tracks 23 are connected to the vacuum suction hole 22 through an internal channel. When the external vacuum pump evacuates air through the vacuum suction hole 22, a negative pressure is formed at the adsorption track 23, so that the silicon lens wafer 5 is flat and stable adsorbed on the surface of the suction cup unit 21, thereby reducing the warping, slippage and local jitter of the silicon lens wafer 5 during the scanning process, so that the silicon lens wafer 5 has a more stable height and positional relationship with the scanning mechanism 4, which facilitates the subsequent acquisition of clear reference images and test images.

[0032] In one specific embodiment, the first direction can be the Y-axis direction (defined as the right-angled sides on the same plane of the supporting mechanism 1, i.e., the first direction and the second direction are perpendicular to each other), and the second direction can be the X-axis direction. The first motion unit 31 includes two drive sub-units arranged parallel to each other along the Y-axis direction and fixed relative to each other on both sides of the base plate. After the two drive sub-units move synchronously, they drive the second motion unit 32 to move along the Y-axis direction as a whole. The second motion unit 32 can be a drive assembly arranged along the X-axis direction, with its two ends respectively connected to the movable part of the first motion unit 31. The scanning mechanism 4 is fixedly connected to the movable part of the second motion unit 32 and can move along the X-axis direction with the second motion unit 32. It is understood that the above-mentioned drive sub-units can be implemented by using a servo motor in conjunction with a lead screw, synchronous belt, linear guide rail or slide table structure, or other drive mechanisms that can output stable linear displacement. Through the cooperation of the first motion unit 31 and the second motion unit 32, the scanning mechanism 4 can form a two-dimensional scanning range above the silicon lens wafer 5, thereby covering the silicon lens units at different positions on the wafer.

[0033] It is also understandable that the first motion unit 31 and the second motion unit 32 have similar structures, and a limiting unit 33 is provided on it. The limiting unit 33 includes a first limiter 331, a second limiter 332, and a limiting piece 333. The first limiter 331 and the second limiter 332 are respectively a negative direction limiter and a positive direction limiter, and are respectively provided at both ends of the motion unit. The limiting piece 333 is provided on the movable part of the motion unit. The first limiter 331 and the second limiter 332 are both normally open photoelectric switches. When the motor in the motion unit drives the limiting piece 333 to the corresponding limiter position, the light is blocked, the limiter outputs a closing signal, and the motor immediately stops moving to prevent the motion mechanism 3 from overtravel, derailing, or colliding.

[0034] In one specific embodiment, the scanning unit 42 includes a camera subunit 421 and a lens subunit 422 in hardware. The camera subunit 421 is preferably a high-resolution CMOS industrial camera, and the lens subunit 422 is connected to the industrial camera through a C interface or other lens interface, such as a telecentric lens. This can reduce the influence of object distance changes and perspective errors on the detection size, so that the silicon lens units at different positions have a more stable imaging ratio in the field of view. The illumination unit 41 is disposed above and / or below the silicon lens wafer 5 to provide illumination conditions according to different defect types. The illumination unit 41 (first illumination source 411) disposed above the silicon lens wafer 5 is preferably a ring-shaped shadowless light source or a ring-shaped shadowless blue light source to enhance the display effect of surface defects such as scratches and pits on the silicon lens surface. The illumination unit 41 (second illumination source 412) disposed below the silicon lens wafer 5 is preferably a planar backlight source or a planar white light source to enhance the contrast of contour defects such as chipped edges and notches by utilizing the light transmission characteristics of the silicon lens. The illumination unit 41 is used in conjunction with the scanning unit 42 so that the same detection device can meet the image acquisition needs of surface defects and edge contour defects of the silicon lens wafer 5.

[0035] During the inspection process, the equipment first acquires reference images corresponding to multiple preset reference mark positions on the silicon lens wafer 5 through the scanning mechanism 4. The reference images contain the wafer reference marks and the silicon lens units within the field of view of the reference marks. The equipment extracts the actual image position of the reference marks and the center position of the lens based on the reference images, and establishes an initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system by combining the physical coordinates of the motion mechanism 3 when acquiring the corresponding reference images. Further, the equipment constructs a two-dimensional coordinate mapping relationship between the wafer map coordinates and the target position of the motion mechanism 3 based on this initial correspondence, so that the arrangement position of the silicon lens units represented in rows and columns in the wafer map can be converted into the target position that the motion mechanism 3 can execute. Based on the above structure, the equipment no longer relies solely on manual alignment or a single mechanical origin for point-by-point scanning, but can generate a scanning path covering the entire wafer based on the arrangement information of the silicon lens units on the wafer.

[0036] It is understood that the suction hole 22 is a suction port for connecting to the vacuum suction air path. One end of it is preferably located at the center of the upper surface of the suction cup unit 21, and the other end is located on the side wall of the suction cup unit 21 and connected to the adsorption track 23 on the upper surface through an internal channel. The adsorption track 23 is preferably a groove structure formed on the upper surface of the suction cup unit 21. The depth and width of the groove are such that a stable negative pressure channel can be formed without affecting the flatness of the wafer support.

[0037] In one specific embodiment, several adsorption tracks 23 are arranged in mutually intersecting directions, such as forming a grid-like track with perpendicular intersections along the first and second directions. Alternatively, they can form a ring-shaped, radial, or spider web-like interconnected track structure according to the actual size and force distribution of the silicon lens wafer 5. Through these interconnected adsorption tracks 23, the negative pressure generated by the vacuum pump during gas extraction can be transmitted more evenly to multiple areas at the bottom of the wafer, thereby forming a continuous or nearly continuous adsorption area on the upper surface of the suction cup unit 21.

[0038] It should be noted that the first illumination source 411 preferably adopts a ring-shaped shadowless blue light source, which can be fixed below the telecentric lens and moves synchronously with the industrial camera and the telecentric lens under the action of the motion mechanism 3, thereby maintaining a relatively consistent illumination angle and brightness in different target scanning areas. The second illumination source 412 is preferably set below the silicon lens wafer 5 and located in the light-transmitting mounting area corresponding to the suction cup unit 21. In specific implementation, the second illumination source 412 preferably adopts a planar white light source, which, together with the transparent acrylic suction cup, forms bottom transmission illumination, so that the light is transmitted to the silicon lens wafer 5 through the suction cup unit 21.

[0039] Since silicon lenses have certain light transmission characteristics, bottom transmission illumination can enhance the contrast between contour defects such as chipped edges, notches, and pits and normal areas. The first illumination source 411 and the second illumination source 412 provide supplementary illumination to the silicon lens wafer 5 from the top reflection illumination and the bottom transmission illumination, respectively, so that the scanning unit 42 can acquire detection images suitable for different defect types in the same detection device, thereby improving the stability of subsequent defect feature extraction and good product determination.

[0040] In this embodiment, the first motion unit 31 includes two driving sub-units arranged opposite to each other, the two ends of the second motion unit 32 are respectively connected to the movable parts of the two driving sub-units, and the scanning mechanism 4 is connected to the movable part of the second motion unit 32 so that the scanning mechanism 4 performs two-dimensional scanning on the silicon lens wafer 5.

[0041] In one specific embodiment, the two drive subunits are formed by using servo motors in conjunction with linear slides, lead screws, or synchronous belts. Their movable parts are used to support the two ends of the second motion unit 32. To ensure that the second motion unit 32 does not deviate during movement, the two drive subunits preferably move synchronously. For example, a motor controller outputs synchronous control signals to the two drive subunits, and they are driven separately by different drivers, so that the two ends of the second motion unit 32 maintain a consistent displacement in the first direction.

[0042] Example 2: As attached Figure 7 As shown, this embodiment provides a method for detecting appearance defects in silicon lenses, including the following steps: The process involves acquiring reference mark images of multiple preset reference mark positions on the silicon lens wafer 5; obtaining image positioning information within the reference mark field of view and the physical coordinates of the motion mechanism 3 when acquiring the corresponding reference mark image based on the reference mark images; wherein the image positioning information includes the actual image position of the reference mark and the center position of the lens, and a plurality of silicon lens units are distributed on the silicon lens wafer 5. Based on the image positioning information and the physical coordinates, the initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system is obtained; and based on the initial correspondence, a two-dimensional coordinate mapping model between the wafer map coordinate system and the motion mechanism coordinate system is constructed. The location of the scanning start area and the product distribution of the silicon lens wafer 5 are obtained. Based on the location of the scanning start area, the product distribution of the silicon lens wafer 5, and the two-dimensional coordinate mapping model, the scanning path covering the silicon lens wafer 5 is obtained. Based on the scanning path and the two-dimensional coordinate mapping model, the theoretical target position corresponding to each target scanning area is obtained; based on the theoretical target position, the motion mechanism 3 is controlled to drive the scanning unit 42 to move sequentially to each target scanning area along the scanning path, and the corresponding target scanning area image is obtained; Based on the target scanning area image, obtain the product image corresponding to each silicon lens unit, obtain the defect feature parameters based on the product image, and obtain the silicon lens appearance defect detection result based on the defect feature parameters.

[0043] This embodiment provides a method for detecting appearance defects of silicon lenses, and is implemented based on the aforementioned detection equipment for appearance defects of silicon lenses. Before the detection, the method further includes a step of fixing the silicon lens wafer 5, specifically including: placing the silicon lens wafer 5 to be tested on the suction cup unit 21, starting the vacuum suction device (vacuum pump), so that the wafer is flat and stably fixed at the detection station by the adsorption track 23 on the surface of the suction cup unit 21.

[0044] After the silicon lens wafer 5 under test is fixed to the fixing mechanism 2, the motion mechanism 3 drives the scanning mechanism 4 to move above the silicon lens wafer 5. The scanning unit 42 sequentially acquires the reference mark image and the target scanning area image. By using the reference mark on the wafer and the center position of the lens in the field of view of the reference mark, a correspondence is established between the wafer map coordinates and the coordinates of the motion mechanism 3. Then, based on this correspondence, a scanning path covering the entire wafer is planned. During the scanning process, the target platform coordinates are corrected by re-anchoring the scanning area, so that subsequent defect identification is based on stable positioning and imaging.

[0045] Specifically, the preset reference mark position is used to guide the scanning unit 42 into the area where the reference mark is located. The reference mark image contains the reference mark and the adjacent silicon lens units within the field of view. During image processing, the actual image position of the reference mark and the lens center position are extracted from the reference mark image, and the physical coordinates of the motion mechanism 3 in the first and second directions are recorded simultaneously when the reference mark image is acquired. The actual image position of the reference mark is used to confirm the true imaging position of the wafer reference point in the image, the lens center position is used to further reflect the actual arrangement state of the silicon lens array near the reference mark, and the physical coordinates of the motion mechanism 3 are used to mark the true arrival position of the scanning unit 42 in the device coordinate system.

[0046] After obtaining the image positioning information and physical coordinates of the motion mechanism 3 corresponding to multiple reference marker positions, the initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system is obtained based on the detection results of these reference points. A two-dimensional coordinate mapping model between the wafer map coordinates and the motion mechanism 3 coordinates is then constructed based on this initial correspondence. Through this two-dimensional coordinate mapping model, the row and column indices in the wafer map can be converted into the target positions that the motion mechanism 3 needs to execute. This eliminates the need for subsequent detection to rely on manual point-by-point adjustments or simple mechanical step-by-step calculations. Instead, the product layout on the silicon lens wafer 5 is directly converted into scan positions that the equipment can execute, providing a unified coordinate basis for automatic scanning of the entire wafer.

[0047] After completing the two-dimensional coordinate mapping, the position of the scanning start area and the product distribution of the silicon lens wafer 5 are obtained. A scanning path covering the silicon lens wafer 5 is generated using the two-dimensional coordinate mapping model. The product distribution refers to the row and column arrangement of the silicon lens units to be inspected on the wafer map. The scanning start area is used to determine the starting position of the full-wafer scan, while the target scanning area corresponds to the local product area covered by the scanning unit 42 in a single imaging operation, such as the area containing three adjacent rows and three columns of silicon lens units. The control system divides the silicon lens units to be inspected on the wafer into multiple target scanning areas based on the product distribution and arranges these areas in a row-by-row reciprocating manner, forming a scanning path covering the entire wafer. This effectively reduces the invalid backhaul of the motion mechanism 3 during the scanning process, allowing the camera to complete full-wafer coverage imaging along a continuous path, thus improving the efficiency of wafer-level inspection.

[0048] During scanning, the theoretical target position corresponding to each target scanning area is obtained according to the scanning path and the two-dimensional coordinate mapping model. The motion mechanism 3 is controlled to drive the scanning unit 42 to move sequentially to each target scanning area along the scanning path and acquire the corresponding target scanning area image. The first motion unit 31 and the second motion unit 32 in the motion mechanism 3 provide displacement in two directions, enabling the scanning unit 42 to reach different target scanning areas. The illumination unit 41 provides corresponding illumination conditions when the scanning unit 42 acquires images to ensure that surface defects and edge defects of the silicon lens can form identifiable grayscale or contour differences in the image.

[0049] In addition, to suppress the influence of the cumulative error of the motion mechanism 3 on the imaging position of the subsequent target scanning area, the following steps are included after acquiring the corresponding target scanning area image: During the scanning process, at predetermined intervals, a target scanning area containing the central product is selected from each target scanning area as a re-anchoring scanning area; based on the target scanning area image corresponding to the re-anchoring scanning area, the actual image center position of the central product is obtained; based on the actual image center position and the current field-of-view image center position of the scanning unit 42, the pixel deviation is obtained; based on the pixel deviation, the corresponding platform compensation amount is obtained; based on the platform compensation amount, the target platform coordinates corresponding to the current target scanning area or subsequent target scanning areas are corrected to complete the online position calibration.

[0050] Understandably, after obtaining the corresponding platform compensation amount based on the pixel deviation, the control system corrects the target platform coordinates corresponding to the current target scanning area or the subsequent target scanning area to complete online position calibration. Through this process, the scanning path can be corrected according to the actual image feedback during execution, reducing the accumulation of positioning deviations caused by long-distance scanning, and ensuring that the imaging position of the subsequent silicon lens unit in the image remains relatively consistent.

[0051] After completing the acquisition of the target scanning area image, the product image corresponding to each silicon lens unit is obtained based on the target scanning area image. Specifically, based on the product distribution of the silicon lens wafer 5 and the position of the target scanning area in the wafer map, the silicon lens unit contained in each target scanning area image is determined. Then, the product image is cropped or separated according to the position of the corresponding silicon lens unit in the image, and the product image is associated with the row and column position of the silicon lens unit in the wafer map coordinate system. This allows the defect results identified later to be written back to the specific silicon lens unit, which is beneficial for forming the defect distribution results of the entire wafer.

[0052] Finally, defect feature parameters are obtained from the product images, and the appearance defect detection results of the silicon lens are obtained based on these parameters. The image processing system can extract the defect area from each product image and obtain feature parameters such as area, aspect ratio, and grayscale contrast based on the defect area. For silicon lens products, pits and dents usually appear as local grayscale abnormal areas, scratches usually appear as thin, elongated line areas, and edge chipping usually appears at the lens edge and forms contour defects.

[0053] Based on the above defect characteristic parameters, the appearance defect type is identified and the corresponding defect size is determined. Then, the corresponding silicon lens unit is judged as good or bad, and finally the appearance defect detection result of the silicon lens is formed. Through the above method, the reference mark alignment, two-dimensional coordinate mapping, target scanning area planning, online re-anchoring and defect identification form a continuous cooperation, so that the entire silicon lens wafer 5 can maintain a relatively stable positioning relationship during the automatic scanning process, and the defect detection result is accurately corresponding to the specific silicon lens unit position.

[0054] In this embodiment, the step of obtaining image positioning information within the field of view of the reference marker and the physical coordinates of the motion mechanism 3 when obtaining the corresponding reference marker image includes the following sub-steps: The reference marker image is subjected to reference marker recognition to obtain the actual image position of the reference marker; The silicon lens unit in the reference mark image is centrally located to obtain the center position of the lens within the field of view of the reference mark; When the scanning unit 42 acquires the corresponding reference mark image, the physical coordinates of the motion mechanism 3 in the first and second directions are obtained. The actual image position of the reference mark, the center position of the lens, and the physical coordinates of the motion mechanism 3 are correlated to form reference alignment data.

[0055] It should be noted that the actual image position of the reference mark is used to characterize the true imaging position of the reference mark in the field of view of the scanning unit 42, the lens center position is used to characterize the actual arrangement of silicon lens units near the reference mark, and the physical coordinates of the motion mechanism 3 in the first and second directions are used to characterize the device position corresponding to the scanning unit 42 when acquiring the image of the reference mark. By associating the above three types of data, a correspondence is established between the positioning result in the image coordinates and the actual position in the coordinates of the motion mechanism 3, providing a data basis for subsequently obtaining the initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system, thereby reducing coordinate mapping errors caused by wafer placement deviations or alignment deviations of the scanning unit 42.

[0056] In this embodiment, the step of obtaining the initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system based on the image positioning information and the physical coordinates, and constructing a two-dimensional coordinate mapping model between the wafer map coordinates and the motion mechanism coordinates based on the initial correspondence, includes the following steps: Based on multiple reference alignment data, obtain the row and column positions of multiple reference points in the wafer map coordinate system and their physical coordinates in the motion mechanism coordinate system; Based on the row and column positions of multiple reference points in the wafer map coordinate system and their physical coordinates in the motion mechanism coordinate system, the initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system is obtained. Based on the initial correspondence, determine the reference row and column positions of the reference reference mark or reference silicon lens unit in the wafer map coordinate system and the reference physical coordinates in the motion mechanism coordinate system; Based on the reference alignment data corresponding to multiple reference points, obtain the row stepping relationship when the wafer map coordinate system increases along the row direction and the column stepping relationship when it increases along the column direction. Based on the reference row and column positions, reference physical coordinates, row stepping relationships, and column stepping relationships, a two-dimensional coordinate mapping model is constructed between the wafer map coordinates and the motion mechanism coordinates.

[0057] It is understandable that the row and column positions of the reference point in the wafer map coordinate system represent the theoretical position of the reference point in the product layout of the silicon lens wafer 5, and the physical coordinates of the reference point in the motion mechanism coordinate system represent the device position when the scanning unit 42 actually acquires the image of the reference point. Through the correspondence between multiple reference points, the overall positional relationship of the wafer relative to the motion mechanism 3 after it is placed on the fixed mechanism 2 is effectively determined, thereby obtaining the initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system.

[0058] After obtaining the initial correspondence, a reference datum marker or reference silicon lens unit is selected as the reference object for coordinate transformation, and its reference row and column positions in the wafer map coordinate system and its reference physical coordinates in the motion mechanism coordinate system are determined. The reference datum marker or reference silicon lens unit is equivalent to the reference point for subsequent coordinate transformation. The row step relationship is used to represent the corresponding change in the target position of motion mechanism 3 when the wafer map coordinates change along the row direction, and the column step relationship is used to represent the corresponding change in the target position of motion mechanism 3 when the wafer map coordinates change along the column direction.

[0059] Therefore, the reference row and column positions, reference physical coordinates, row stepping relationship and column stepping relationship together constitute a two-dimensional coordinate mapping model, which enables any row and column position of the silicon lens unit to be detected in the wafer map to be converted into a target position that can be executed by the motion mechanism 3.

[0060] In this embodiment, obtaining the position of the scanning start area and the product distribution of the silicon lens wafer 5, and obtaining the scanning path covering the silicon lens wafer 5 based on the position of the scanning start area, the product distribution of the silicon lens wafer 5, and the two-dimensional coordinate mapping model, includes the following steps: Based on the product distribution of silicon lens wafer 5, the row and column distribution of the silicon lens unit to be tested in the wafer map coordinate system is obtained. Determine the starting target scanning area of ​​the scanning path based on the location of the scanning start area; Based on the row and column distribution of the silicon lens unit to be tested in the wafer map coordinate system, adjacent multi-row and multi-column silicon lens units to be tested are divided into the same target scanning area. Based on the two-dimensional coordinate mapping model, the theoretical target positions corresponding to each target scanning area are obtained; The target scanning areas are arranged in a row-by-row manner to obtain the scanning path covering the silicon lens wafer 5.

[0061] It should be noted that the product distribution of silicon lens wafer 5 is used to represent the effective row and column range of the silicon lens unit to be tested in the wafer map coordinate system. The position of the scanning start area is used to determine the target scanning area first entered when scanning the entire wafer. After determining the row and column distribution of the silicon lens unit to be tested according to the product distribution, adjacent multi-row and multi-column silicon lens units to be tested are divided into the same target scanning area, so that the scanning unit 42 covers multiple silicon lens units in one imaging process. For example, if the field of view of the scanning unit 42 allows, adjacent three rows and three columns of silicon lens units can be divided into one target scanning area to reduce the number of movements required for individual positioning and individual imaging of each silicon lens unit.

[0062] After determining the target scanning areas, the positions of each target scanning area in the wafer map coordinate system are converted into theoretical target positions in the motion mechanism coordinate system based on the two-dimensional coordinate mapping model. These theoretical target positions are used to instruct the motion mechanism 3 to move the scanning unit 42 to the corresponding area, enabling the scanning unit 42 to align with the corresponding silicon lens unit group for image acquisition. Subsequently, the target scanning areas are arranged in a row-by-row reciprocating manner; that is, after completing one row of target scanning areas, scanning continues in the reverse direction in the adjacent next row to reduce the idle travel and backlash of the motion mechanism 3. Through this method, while ensuring that the target scanning area covers the distribution of the product to be inspected, the scanning continuity of the entire silicon lens wafer 5 is improved.

[0063] In this embodiment, the step of controlling the motion mechanism 3 to drive the scanning unit 42 to move sequentially along the scanning path to each target scanning area according to the theoretical target position, and acquiring the corresponding target scanning area image, includes the following sub-steps: Based on the theoretical target position corresponding to each target scanning area, motion control commands are generated; Based on the lighting requirements corresponding to each target scanning area, generate lighting control commands; According to the motion control command, the motion mechanism 3 drives the scanning unit 42 to move sequentially to each target scanning area along the scanning path; According to the lighting control command, the lighting unit 41 is controlled to illuminate the target scanning area currently corresponding to the scanning unit 42; The control scanning unit 42 acquires the corresponding target scanning area image.

[0064] It is understood that the motion control command includes the moving direction, moving distance and stopping position corresponding to the first motion unit 31 and the second motion unit 32 respectively, so that the motion mechanism 3 can drive the scanning unit 42 to reach each target scanning area sequentially along the scanning path. Since the scanning mechanism 4 is connected to the moving part of the second motion unit 32, after the first motion unit 31 and the second motion unit 32 cooperate to move, the scanning unit 42 completes the two-dimensional position switching above the silicon lens wafer 5.

[0065] When the scanning unit 42 reaches the current target scanning area, it generates an illumination control command based on the illumination requirements of the current target scanning area and controls the illumination unit 41 to illuminate the target scanning area. Specifically, for images that need to highlight defects such as scratches and pits on the silicon lens surface, the first illumination source 411 positioned above the silicon lens wafer 5 is controlled to provide supplementary illumination; for images that need to enhance contour features such as chipped edges and notches, the second illumination source 412 positioned below the silicon lens wafer 5 is controlled to provide supplementary illumination. The illumination unit 41 and the scanning unit 42 work together to ensure that the target scanning area has relatively stable brightness and contrast during image acquisition.

[0066] After the motion mechanism 3 reaches the corresponding theoretical target position and completes the illumination, the scanning unit 42 acquires the current target scanning area image. The target scanning area image contains one or more silicon lens units within the current field of view. By executing motion control, illumination control and image acquisition one by one according to the target scanning area, the scanning unit 42 continuously acquires the detection image of the entire silicon lens wafer 5 according to the predetermined path, and reduces the image quality fluctuation caused by the position not being in place or the illumination being unstable.

[0067] In this embodiment, obtaining the corresponding platform compensation amount based on the pixel deviation includes the following sub-steps: Obtain the correspondence between the pre-calibrated image coordinates and the coordinates of the motion mechanism 3; Based on the pixel deviation and the correspondence between the image coordinates and the coordinates of the motion mechanism 3, the position offset of the pixel deviation in the coordinate system of the motion mechanism is obtained; The position offset is used as the platform compensation amount.

[0068] Since pixel deviation itself reflects the imaging offset of the central product in the target scanning area image, it belongs to the deviation amount under the image coordinate system and cannot be directly used as the correction amount of motion mechanism 3. Therefore, when obtaining the platform compensation amount, the correspondence between the pre-calibrated image coordinates and the coordinates of motion mechanism 3 is first called. This correspondence can be obtained during equipment debugging or reference alignment and is used to represent the conversion relationship between the pixel position change in the image and the actual displacement of motion mechanism 3 in the first direction and the second direction.

[0069] Furthermore, based on the pixel deviation and the correspondence between the image coordinates and the coordinates of the motion mechanism 3, the pixel offset of the central product relative to the center of the current field of view image can be converted into the position offset in the coordinate system of the motion mechanism. This reflects the direction and distance that the current position of the scanning unit 42 needs to be compensated for relative to the target platform coordinates. After using this position offset as the platform compensation amount, the control system can correct the target platform coordinates corresponding to the current target scanning area or the subsequent target scanning area, so that the scanning unit 42 is closer to the actual center position of the silicon lens unit in the subsequent scanning, thereby reducing the impact of the accumulation of position error on the consistency of image acquisition during continuous scanning.

[0070] In this embodiment, the step of obtaining product images corresponding to each silicon lens unit based on the target scanning area image, obtaining defect feature parameters based on the product images, and obtaining silicon lens appearance defect detection results based on the defect feature parameters includes the following sub-steps: Based on the product distribution of silicon lens wafer 5, determine the silicon lens units contained in the image of each target scanning area. Based on the position of the silicon lens unit in the image of each target scanning area, obtain the product image of the corresponding silicon lens unit; Associate the product image with the row and column positions of the corresponding silicon lens unit in the wafer map coordinate system; Defect regions are extracted from the product image to obtain the defect regions; Based on the defect region, defect feature parameters are obtained, including area features, aspect ratio features, and grayscale contrast features. Based on the defect feature parameters, the appearance defect type and the corresponding defect size are obtained, wherein the appearance defect type includes at least one of pitting, dents, chipping, and scratches; Based on the type and size of the appearance defect, obtain the good product or defective product judgment result of the corresponding silicon lens unit; Based on the good or bad product determination results of each silicon lens unit, obtain the appearance defect detection results of the silicon lens.

[0071] It should be noted that after obtaining the product image, suspected abnormal areas are extracted to obtain defect areas. Defect areas can be determined by grayscale differences, edge contour changes, or semantic segmentation results. Subsequently, area features, aspect ratio features, and grayscale contrast features are calculated based on the defect areas. The area feature reflects the image range occupied by the defect, the aspect ratio feature distinguishes between linear and point defects, and the grayscale contrast feature reflects the brightness difference between the defect area and the surrounding normal area. Based on these features, at least one type of appearance defect, such as pitting, dents, chipping, and scratches, can be identified, and the corresponding defect size can be determined.

[0072] Furthermore, based on the type and size of the appearance defects, the corresponding silicon lens unit is judged as either a good or a bad product. For example, when the defect size or number of defects exceeds the preset judgment conditions, the silicon lens unit can be judged as a bad product; when no defects exceeding the judgment conditions are identified, the silicon lens unit can be judged as a good product. Finally, the judgment results of each silicon lens unit are summarized to form the silicon lens appearance defect detection results, so that the detection results can simultaneously reflect the defect type, defect size, and wafer row and column position where the defect is located, thereby facilitating the subsequent generation of defect distribution results and product traceability.

[0073] In one specific implementation, this embodiment illustrates the specific process of detecting appearance defects in silicon lenses in conjunction with the aforementioned detection equipment and methods.

[0074] Its specific process includes: Step 1, wafer loading and vacuum adsorption fixation: Place the silicon lens wafer 5 to be tested on the suction cup unit 21, start the vacuum pump, and fix the wafer to the testing station by the adsorption track 23 on the surface of the suction cup unit 21.

[0075] Step 2, datum mark alignment and initial coordinate calibration: Control the motion mechanism 3 mentioned above to drive the industrial camera to move sequentially to the area of ​​multiple preset datum marks on the wafer, acquire corresponding images, extract the corresponding position of the mark and the center position of the lens within the field of view of each datum mark, and obtain the correspondence between the image coordinates and the platform coordinates; Based on the detection results of multiple datum points, determine the initial correspondence between the wafer map coordinate system and the physical coordinate system of the motion mechanism 3.

[0076] Step 3, Two-dimensional coordinate mapping model establishment and dynamic compensation: Based on the coordinates of multiple reference points obtained in Step 2, a two-dimensional mapping model is established between the row and column coordinates of the wafer map and the X and Y physical coordinates of the motion platform. The mathematical expression of the two-dimensional mapping model is shown in the following formula (1): Equation (1): ; In the above formula (1), This represents the target row and column index coordinates of the current target lens unit in the wafer mapping diagram, where r is the row index and c is the column index; (x,y) represents the reference row and column index coordinates of the pre-selected reference alignment mark or reference lens unit in the wafer mapping diagram; (x,y) represents the calculated physical target position coordinates when the motion mechanism 3 carries the target lens unit to the center of the camera's field of view. () indicates the actual physical position coordinates of the reference alignment mark or reference lens unit in the motion mechanism coordinate system; This represents the two-dimensional step vector in the motion mechanism coordinate system when the wafer map coordinate system increments by one unit index along the row direction; This represents the two-dimensional step vector in the motion mechanism coordinate system when the wafer map coordinate system increments by one unit index along the column direction.

[0077] During the subsequent alignment process, the controller dynamically corrects the row and column step vectors based on the newly added valid alignment samples to compensate for platform assembly errors, drift errors, and cumulative motion errors, thereby improving the positioning accuracy during long-distance scanning.

[0078] Step 4, Scan Path Planning: Based on the wafer map layout, the baseline starting scan position, and the distribution of the target products to be inspected, a scan path covering the entire wafer is generated. The scan path is planned in 3×3 units and uses a row-by-row S-shaped scanning method to ensure that each exposure covers an adjacent 3×3 product area, thereby reducing the number of movements and improving inspection efficiency, as shown in Equations (2) and (3): Equation (2): ; Equation (3): ; in, This represents the two-dimensional pixel coordinates in the image space acquired by the industrial camera, where For pixel column coordinates, These are the row coordinates of the pixels; This represents the physical coordinates of the corresponding motion platform; This represents the two-dimensional affine transformation matrix from the pre-calibrated camera pixel coordinate system to the motion platform physical coordinate system; This indicates the actual geometric center pixel coordinates of the anchored target lens unit detected during the current online re-anchoring phase within the current field of view; Represents the geometric and physical center pixel coordinates of the field of view of an industrial camera image; This represents the calculated real-time physical position feedback compensation amount of the motion platform.

[0079] Step 5, Image Acquisition and Online Re-anchoring: The motion mechanism 3 is controlled to move sequentially to the theoretical position of each target scanning area and acquire images. To suppress the cumulative error of the platform, during the scanning process, the center product is selected as the re-anchoring scanning area at predetermined intervals. The actual image center position of the center product in the scanning area is extracted and compared with the current field of view image center to calculate the platform compensation amount corresponding to the pixel deviation. The controller corrects the target platform coordinates of the current scanning area or subsequent scanning areas accordingly to complete the online position calibration.

[0080] Step 6, Defect Identification and Judgment: The image processing system first extracts the complete lens from the acquired 3×3 field-of-view image to obtain the ROI image of a single silicon lens; then, it performs circular boundary localization, central region cropping, scale normalization, and masking on the single lens ROI; finally, it inputs the processed central region image into a deep learning network weight model optimized for minor defects.

[0081] The network outputs the good / defect probability of a single silicon lens, automatically determines it based on a preset threshold, and writes the determination result back to the wafer map for subsequent sorting and rejection. In addition, as an extended implementation, a special defect detection module such as edge chipping can be added to the edge area.

[0082] Step 7, Result Output and Subsequent Processing: Based on the defect identification results, the system automatically determines all silicon lens products as good or bad, generating a defect distribution map containing the precise location coordinates, defect type, and size of each defective product. This distribution map is then exported and imported into the downstream automatic unloading machine to achieve automated rejection of defective products. Good products then enter the subsequent packaging process to complete all inspection operations.

[0083] In one specific implementation, for step 6 above, the image processing in this embodiment utilizes a deep learning network weight model pre-trained with a large number of 10-micron-level defect samples to perform pixel-level semantic segmentation analysis on each product image.

[0084] Specifically, it prioritizes using the category-weighted cross-entropy loss function to train the defect recognition network, in order to address the problems of a small number of 10-micron-level extremely small defect samples, an imbalance in the distribution of positive and negative samples, and the fact that minority class defect features are easily overwhelmed by majority class good product features.

[0085] More specifically, let the number of good samples be... The number of defective samples is The total sample size is The two types of weights are defined as follows: ; For any sample, its weighted cross-entropy loss can be expressed as: ; Where y is the true category of the sample; This represents the probability that the network predicts the true category. This represents the training weights corresponding to the good product category; This represents the training weights corresponding to the defective product category; Wy represents the category weight corresponding to the true category y; This represents the weighted cross-entropy loss value corresponding to the current sample.

[0086] By assigning higher loss weights to minority class defect samples, the network's ability to respond to minor defects can be effectively improved, and the dominant role of majority class samples in gradient updates during training can be reduced. After the network training is completed, the judgment threshold is searched and optimized in conjunction with the target defect recall rate, thereby prioritizing the detection rate of defects in actual detection and reducing the risk of missed detection.

[0087] Example 3: As attached Figure 8 As shown, this embodiment provides a detection system for appearance defects in silicon lenses, the system comprising: The positioning information acquisition module is used to acquire reference mark images of multiple preset reference mark positions of the silicon lens wafer 5; based on the reference mark images, it acquires image positioning information within the reference mark field of view and the physical coordinates of the motion mechanism 3 when acquiring the corresponding reference mark image; wherein, the image positioning information includes the actual image position of the reference mark and the center position of the lens, and a plurality of silicon lens units are distributed on the silicon lens wafer 5; The coordinate mapping model construction module is used to obtain the initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system based on the image positioning information and the physical coordinates; and to construct a two-dimensional coordinate mapping model between the wafer map coordinate system and the motion mechanism coordinate system based on the initial correspondence. The scan path acquisition module is used to acquire the position of the scan start area and the product distribution of the silicon lens wafer 5. Based on the position of the scan start area, the product distribution of the silicon lens wafer 5 and the two-dimensional coordinate mapping model, the scan path covering the silicon lens wafer 5 is acquired. The product image acquisition module is used to acquire the theoretical target position corresponding to each target scanning area according to the scanning path and the two-dimensional coordinate mapping model; according to the theoretical target position, the motion mechanism 3 is controlled to drive the scanning unit 42 to move sequentially to each target scanning area along the scanning path, and acquire the corresponding target scanning area image. The detection result acquisition module is used to acquire product images corresponding to each silicon lens unit based on the target scanning area image, acquire defect feature parameters based on the product images, and acquire silicon lens appearance defect detection results based on the defect feature parameters.

[0088] It should be noted that the detection system in this embodiment is based on the aforementioned detection equipment and detection method. Each module can be completed by the control system, image processing system, motion control component and scanning mechanism 4. The positioning information acquisition module is mainly used to obtain the reference mark image, the actual image position of the reference mark, the center position of the lens and the physical coordinates of the motion mechanism 3, so as to provide basic data for the subsequent establishment of the correspondence between the wafer map coordinate system and the motion mechanism coordinate system. The coordinate mapping model construction module forms a two-dimensional coordinate mapping model on the basis of the above basic data, so that the row and column positions of the silicon lens unit in the wafer map can be converted into the target positions that the motion mechanism 3 can execute.

[0089] The scanning path acquisition module generates a scanning path covering the silicon lens wafer 5 based on the scanning start area, product distribution, and two-dimensional coordinate mapping model. The scanning image acquisition module then controls the motion mechanism 3 to move the scanning unit 42 sequentially to each target scanning area according to the scanning path and acquires the image of the target scanning area.

[0090] The product image acquisition module is used to separate the product images corresponding to each silicon lens unit from the target scanning area image and keep the product images corresponding to the row and column positions in the wafer map coordinate system. The detection result acquisition module further extracts defect feature parameters based on the product images, identifies appearance defect types such as pits, dents, chipping, and scratches, and generates corresponding good product judgment results or defective product judgment results, so that the whole-wafer scanning, position correction and defect detection result output of silicon lens wafer 5 can be continuously completed in the same system.

[0091] In one specific implementation, the system further includes a position calibration module, which is used to select a re-anchoring scanning area during continuous scanning and obtain a platform compensation amount based on the pixel deviation between the actual image center position of the central product and the current field-of-view image center position, thereby correcting the target platform coordinates of the current target scanning area or the subsequent target scanning area and reducing the cumulative positioning deviation during the scanning process.

[0092] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A device for detecting surface defects in silicon lenses, characterized in that, include: Bearing mechanism A fixing mechanism is disposed on a support mechanism and is used to fix a silicon lens wafer. A motion mechanism is disposed on a support mechanism, the motion mechanism includes a first motion unit that moves along a first direction and a second motion unit that moves along a second direction, the second motion unit being disposed in the movable part of the first motion unit; A scanning mechanism is connected to the second motion unit. The scanning mechanism includes an illumination unit and a scanning unit, with the scanning unit positioned above the silicon lens wafer and the illumination unit positioned above and / or below the silicon lens wafer.

2. The detection device for appearance defects of silicon lenses as described in claim 1, characterized in that, The fixing mechanism includes a suction cup unit. The upper end face of the suction cup unit is provided with suction holes and several adsorption tracks. The adsorption tracks are connected to a vacuum pump through the suction holes. The several adsorption tracks are interconnected to form an adsorption area on the upper end face of the suction cup unit.

3. The detection device for appearance defects of silicon lenses as described in claim 2, characterized in that, The lighting unit includes a first lighting source and a second lighting source. The first lighting source is disposed above the silicon lens wafer, and the second lighting source is disposed below the silicon lens wafer and located inside the suction cup unit.

4. The detection device for appearance defects of silicon lenses as described in claim 1, characterized in that, The first motion unit includes two driving sub-units arranged opposite each other. The two ends of the second motion unit are respectively connected to the movable parts of the two driving sub-units. The scanning mechanism is connected to the movable part of the second motion unit so that the scanning mechanism performs two-dimensional scanning on the silicon lens wafer.

5. A method for detecting appearance defects in silicon lenses, characterized in that, A detection device for appearance defects of a silicon lens according to any one of claims 1 to 4, comprising the following steps: A reference mark image is acquired at multiple preset reference mark positions on a silicon lens wafer; based on the reference mark image, image positioning information within the reference mark field of view and the physical coordinates of the motion mechanism when acquiring the corresponding reference mark image are obtained; wherein, the image positioning information includes the actual image position of the reference mark and the center position of the lens, and a plurality of silicon lens units are distributed on the silicon lens wafer; Based on the image positioning information and the physical coordinates, an initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system is obtained; and based on the initial correspondence, a two-dimensional coordinate mapping model between the wafer map coordinate system and the motion mechanism coordinate system is constructed. The location of the scanning start area and the product distribution of the silicon lens wafer are obtained. Based on the location of the scanning start area, the product distribution of the silicon lens wafer, and the two-dimensional coordinate mapping model, the scanning path covering the silicon lens wafer is obtained. Based on the scanning path and the two-dimensional coordinate mapping model, the theoretical target position corresponding to each target scanning area is obtained; based on the theoretical target position, the motion mechanism is controlled to drive the scanning unit to move sequentially to each target scanning area along the scanning path, and the corresponding target scanning area image is obtained; Based on the target scanning area image, obtain the product image corresponding to each silicon lens unit, obtain the defect feature parameters based on the product image, and obtain the silicon lens appearance defect detection result based on the defect feature parameters.

6. The method for detecting appearance defects in a silicon lens as described in claim 5, characterized in that, After obtaining the corresponding target scan area image, the following steps are included: During the scanning process, at predetermined intervals, a target scanning area containing the central product is selected from each target scanning area as a re-anchoring scanning area; based on the target scanning area image corresponding to the re-anchoring scanning area, the actual image center position of the central product is obtained; based on the actual image center position and the current field-of-view image center position of the scanning unit, the pixel deviation is obtained; based on the pixel deviation, the corresponding platform compensation amount is obtained; based on the platform compensation amount, the target platform coordinates corresponding to the current target scanning area or subsequent target scanning areas are corrected to complete online position calibration.

7. The method for detecting appearance defects in a silicon lens as described in claim 5, characterized in that, The step of obtaining image positioning information within the field of view of the reference marker and the physical coordinates of the motion mechanism when obtaining the corresponding reference marker image includes the following steps: The reference marker image is subjected to reference marker recognition to obtain the actual image position of the reference marker; The silicon lens unit in the reference mark image is centrally located to obtain the center position of the lens within the field of view of the reference mark; When the scanning unit acquires the corresponding reference mark image, the physical coordinates of the motion mechanism in the first and second directions are obtained; The actual image position of the reference mark, the center position of the lens, and the physical coordinates of the motion mechanism are correlated to form reference alignment data.

8. The method for detecting appearance defects in a silicon lens as described in claim 7, characterized in that, The initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system is obtained based on the image positioning information and the physical coordinates. Based on the initial correspondence, a two-dimensional coordinate mapping model between the wafer map coordinate system and the motion mechanism coordinate system is constructed, including the following steps; Based on multiple reference alignment data, obtain the row and column positions of multiple reference points in the wafer map coordinate system and their physical coordinates in the motion mechanism coordinate system; Based on the row and column positions of multiple reference points in the wafer map coordinate system and their physical coordinates in the motion mechanism coordinate system, the initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system is obtained. Based on the initial correspondence, determine the reference row and column positions of the reference reference mark or reference silicon lens unit in the wafer map coordinate system and the reference physical coordinates in the motion mechanism coordinate system; Based on the reference alignment data corresponding to multiple reference points, obtain the row stepping relationship when the wafer map coordinate system increases along the row direction and the column stepping relationship when it increases along the column direction. Based on the reference row and column positions, reference physical coordinates, row stepping relationships, and column stepping relationships, a two-dimensional coordinate mapping model between the wafer map coordinate system and the motion mechanism coordinate system is constructed.

9. The method for detecting appearance defects in a silicon lens as described in claim 5, characterized in that, The process of obtaining the location of the scanning start area and the product distribution of the silicon lens wafer, and obtaining the scanning path covering the silicon lens wafer based on the location of the scanning start area, the product distribution of the silicon lens wafer, and the two-dimensional coordinate mapping model, includes the following steps: Based on the product distribution of the silicon lens wafer, obtain the row and column distribution of the silicon lens unit to be tested in the wafer map coordinate system; Determine the starting target scanning area of ​​the scanning path based on the location of the scanning start area; Based on the row and column distribution of the silicon lens unit to be tested in the wafer map coordinate system, adjacent multi-row and multi-column silicon lens units to be tested are divided into the same target scanning area. Based on the two-dimensional coordinate mapping model, the theoretical target positions corresponding to each target scanning area are obtained; The target scanning areas are arranged in a row-by-row manner to obtain the scanning path covering the silicon lens wafer.

10. The method for detecting appearance defects in a silicon lens as described in claim 5, characterized in that, The step of controlling the motion mechanism to move the scanning unit sequentially along the scanning path to each target scanning area according to the theoretical target position, and acquiring the corresponding target scanning area image, includes the following steps: Based on the theoretical target position corresponding to each target scanning area, motion control commands are generated; Based on the lighting requirements corresponding to each target scanning area, generate lighting control commands; According to the motion control command, the motion mechanism is controlled to drive the scanning unit to move sequentially to each target scanning area along the scanning path; According to the lighting control command, the lighting unit is controlled to illuminate the target scanning area currently corresponding to the scanning unit; The control scanning unit acquires the corresponding target scanning area image.

11. The method for detecting appearance defects in a silicon lens as described in claim 6, characterized in that, The step of obtaining the corresponding platform compensation amount based on the pixel deviation includes the following steps: Obtain the correspondence between pre-calibrated image coordinates and motion mechanism coordinates; Based on the pixel deviation and the correspondence between the image coordinates and the motion mechanism coordinates, the position offset of the pixel deviation in the motion mechanism coordinate system is obtained; The position offset is used as the platform compensation amount.

12. The method for detecting appearance defects in a silicon lens as described in claim 5, characterized in that, The process of obtaining product images corresponding to each silicon lens unit based on the target scanning area image, obtaining defect feature parameters based on the product images, and obtaining silicon lens appearance defect detection results based on the defect feature parameters includes the following steps: Based on the product distribution of the silicon lens wafer, determine the silicon lens units contained in the image of each target scanning area; Based on the position of the silicon lens unit in the image of each target scanning area, obtain the product image of the corresponding silicon lens unit; Associate the product image with the row and column positions of the corresponding silicon lens unit in the wafer map coordinate system; Defect regions are extracted from the product image to obtain the defect regions; Based on the defect region, defect feature parameters are obtained, including area features, aspect ratio features, and grayscale contrast features. Based on the defect feature parameters, the appearance defect type and the corresponding defect size are obtained; wherein, the appearance defect type includes at least one of pitting, dents, chipping, and scratches; Based on the type and size of the appearance defect, obtain the good product or defective product judgment result of the corresponding silicon lens unit; Based on the good or bad product determination results of each silicon lens unit, obtain the appearance defect detection results of the silicon lens.

13. A system for detecting surface defects in silicon lenses, characterized in that, The system includes: The positioning information acquisition module is used to acquire reference mark images of multiple preset reference mark positions on the silicon lens wafer; based on the reference mark images, it acquires image positioning information within the reference mark field of view and the physical coordinates of the motion mechanism when acquiring the corresponding reference mark image; wherein, the image positioning information includes the actual image position of the reference mark and the center position of the lens, and the silicon lens wafer is distributed with a plurality of silicon lens units; The coordinate mapping model construction module is used to obtain the initial correspondence between the wafer map coordinate system and the motion mechanism coordinate system based on the image positioning information and the physical coordinates; and to construct a two-dimensional coordinate mapping model between the wafer map coordinate system and the motion mechanism coordinate system based on the initial correspondence. The scan path acquisition module is used to acquire the position of the scan start area and the product distribution of the silicon lens wafer. Based on the position of the scan start area, the product distribution of the silicon lens wafer, and the two-dimensional coordinate mapping model, the scan path covering the silicon lens wafer is acquired. The product image acquisition module is used to obtain the theoretical target position corresponding to each target scanning area according to the scanning path and the two-dimensional coordinate mapping model; according to the theoretical target position, it controls the motion mechanism to drive the scanning unit to move sequentially to each target scanning area along the scanning path and acquire the corresponding target scanning area image. The detection result acquisition module is used to acquire product images corresponding to each silicon lens unit based on the target scanning area image, acquire defect feature parameters based on the product images, and acquire silicon lens appearance defect detection results based on the defect feature parameters.

Images