Transparent part defect detection method and system based on dual-optical-path phase deflection technology

Abstract

The invention discloses a transparent part defect detection method and system based on dual-optical-path phase deflection, and belongs to the technical field of optical detection. The system comprises a camera, a reflection light source, a spectroscope, a backlight light source and a processing unit. The detection method comprises the steps of primary measurement, product overturning, secondary measurement and defect comprehensive analysis and classification. A depth map and a modulation degree map of the surface of a product are obtained through reflective measurement, a phase difference map and a backlight modulation degree map are obtained through transmission-type zero reference surface differential measurement, multi-dimensional data are fused, and accurate detection and classification of concave and convex points, foreign matter, scratches and other defects are achieved according to preset logic. The method can effectively eliminate inherent errors of the system, realizes double-sided full detection without omission, is high in detection precision and accurate in defect classification, and is suitable for defect detection of transparent parts such as multilayer transparent glass.

Description

Technical Field

[0001] The present invention relates to the field of optical detection technology, and in particular to a method and system for detecting inner and outer surface and internal defects of transparent parts such as ultra-thin glass and multi-layer composite glass using phase measurement deflectometry. Background Art

[0002] At present, defect detection of multi-layer transparent glass such as mobile phone cover plates and car screens mainly relies on manual visual inspection or traditional machine vision technology.

[0003] Manual visual inspection: It relies heavily on workers’ experience and has problems such as low efficiency, inconsistent standards, and fatigue leading to missed inspections and false inspections. It cannot meet the needs of large-scale, high-precision production.

[0004] Traditional machine vision uses cameras to capture images of products under normal lighting. This technology has a very low detection rate for subtle defects such as scratches, internal bubbles, inclusions, concave and convex spots, and foreign matter on transparent objects like glass. These defects do not emit or absorb significant amounts of light, but only slightly disrupt the propagation path of transmitted or reflected light. These defects have very low contrast in ordinary images, making them difficult to detect.

[0005] Single-path phase deflectometry: Existing technologies use single-mode reflective or transmissive phase deflectometry for inspection. Reflective methods are primarily sensitive to the sample's front surface morphology and defects; transmissive methods detect the cumulative phase change after light passes through the entire sample. Both methods struggle to effectively separate and locate defects on the front and back surfaces, as well as interlayer foreign matter and scratches within the material, easily leading to confusion regarding defect attribution. Furthermore, inherent geometric errors in the optical system (such as camera lens distortion and uneven light source screens) are directly added to the measurement results, impacting inspection accuracy. Complex and expensive calibration processes are required to eliminate or mitigate these system errors.

[0006] Therefore, how to accurately and efficiently detect and accurately classify the internal and external comprehensive defects of transparent parts, effectively distinguish the sources of defects, and eliminate the system's own errors is a technical problem that needs to be solved urgently in this field. Summary of the Invention

[0007] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a detection method and system that can accurately detect and classify various defects such as concave and convex points, surface foreign matter, interlayer foreign matter, surface and interlayer scratches on transparent parts such as multi-layer transparent glass, and can effectively eliminate the inherent errors of the system and ensure full double-sided inspection without omission.

[0008] The present invention proposes a transparent component defect detection system based on dual-path phase deflectometry, comprising a camera, a reflective light source, a beam splitter, a backlight source, and a processing unit. The camera is positioned directly above the product under test and is used to capture fringe images. The reflective light source projects multi-frequency phase-shifted fringe light. The beam splitter is positioned between the camera and the product under test and reflects the fringe light projected by the reflective light source onto the upper surface of the product under test. The backlight source, a programmable screen, is positioned directly below the product under test and projects multi-frequency phase-shifted fringe light through the product under test. The processing unit is electrically connected to the camera, the reflective light source, and the backlight source and is used to control the timing of the reflected and backlight sources, process the images captured by the camera, and execute a defect detection algorithm. The present invention utilizes a dual-path (reflective and transmissive) design to simultaneously detect surface and internal defects in transparent components. The beam splitter ensures that the reflective and imaging light paths do not conflict, improving optical path utilization. The centralized control of the processing unit ensures accurate measurement timing and consistent data processing, resolving the problem of a single optical path being unable to distinguish defect locations (surface / internal).

[0009] Furthermore, the camera is a 50-80 million pixel industrial camera, preferably a 65 million pixel industrial camera.

[0010] Furthermore, the backlight source is an LCD screen or an OLED screen. The LCD / OLED screen serves as a programmable light source and can flexibly adjust the frequency, phase, and brightness of the multi-frequency phase-shift stripes to meet the detection needs of transparent parts of different thicknesses and materials (such as glass and acrylic).

[0011] In this invention, a reflective light source, a spectroscope, and a camera work together to form a reflective measurement optical path. The reflective light source provides a specific fringe light pattern, the spectroscope redirects the light pattern so that it strikes the surface being measured, and the camera captures the reflected fringe image. Together, these three enable reflective phase deflectometry measurement of the product surface (sides A and B), providing the data foundation for obtaining depth and modulation maps of the surface.

[0012] The backlight source and camera work together to form a transmissive measurement optical path. The striped light projected by the backlight source penetrates the product under test and is captured by the camera. By comparing it with the reference phase when there is no sample, a phase difference map and a backlight modulation map reflecting the comprehensive conditions inside the product and on both sides are obtained, which complement the data obtained by reflective measurement.

[0013] The processing unit, at its core, coordinates with the camera and light source to control the timing of light projection, ensuring accurate and synchronized image acquisition. It processes and analyzes the images captured by the camera, integrating multi-dimensional data from reflective and transmissive measurements, and employs defect classification logic to accurately identify and classify defects. The product flipping step, combined with the two front-to-back measurements, enables the system to inspect both surfaces of the product independently, ensuring a complete, double-sided inspection.

[0014] The present invention also provides a method for detecting defects in transparent parts based on dual-path phase deflectometry, comprising the following steps: First measurement: Start the reflective light source and perform reflective phase deflectometry on surface A of the product under test. The camera captures the reflected fringe image. The processing unit calculates the depth map reflecting the three-dimensional topography of surface A and the reflective modulation map reflecting the fringe contrast based on the captured reflected fringe image. Simultaneously, start the backlight source and project multi-frequency phase-shifted fringe light through the product under test. The camera captures the transmitted fringe image and performs zero-reference plane differential measurement. That is, first collect the reference phase map without a sample, then collect the measured phase map with a sample to obtain the phase difference map. At the same time, the processing unit calculates the backlight modulation map based on the transmitted fringe image. Product flip: After completing the first measurement, flip the product to be measured so that its B side faces the camera; Re-measurement: Repeat the reflective phase deflectometry measurement steps for surface A in the first measurement step, and perform reflective phase deflectometry measurement on surface B of the product. The camera captures the reflected fringe image, and the processing unit obtains the depth map and reflection modulation map of surface B based on the captured reflected fringe image. Comprehensive Defect Analysis and Classification: The processing unit integrates all of the aforementioned measurement data and identifies and classifies defects based on pre-set logic. This system combines dual-sided reflectance and transmission differential measurements to comprehensively inspect defects on both sides of a transparent component, including those within the component. Zero-reference plane differential measurement eliminates inherent system errors. The product flipping step ensures consistent inspection conditions on both sides, preventing omissions during single-sided inspection. Multi-data fusion analysis eliminates misjudgment of defect types (e.g., surface foreign matter or interlayer foreign matter).

[0015] Furthermore, in the reflective phase deflectometry measurement of surface A and surface B, the N-step phase shift method is used to extract the package phase and modulation. The grayscale value Ii(x,y) of the i-th fringe image captured by the camera is expressed as: i (x,y)=A(x,y)+B(x,y)cos[Φ(x,y)+δ i ], where (x, y) is the pixel coordinate, A(x, y) is the background intensity map, B(x, y) is the fringe amplitude map, Φ(x, y) is the phase to be determined, δ i=2πi / N is the phase shift of the i-th step, and N is the total number of steps. The N-step phase shift method effectively suppresses the noise interference of a single image by superimposing the phases of multiple images.

[0016] Furthermore, N is 4, and the δ of the four-step phase shift is i =0,π / 2,π,3π / 2, the wrapped phase diagram Φw(x,y) is simplified to: The four-step phase shift method reduces the number of image acquisitions while ensuring the accuracy of phase calculation.

[0017] Furthermore, the calculation formulas for the fringe amplitude map B(x,y) and the background light intensity map A(x,y) are: , ; The modulation map M(x,y) is defined as the ratio of the fringe amplitude map to the background intensity map, that is, The modulation map can intuitively reflect the degree to which the stripes are disturbed by defects (such as blurred stripes caused by foreign matter), and can be used together with the depth map to accurately distinguish defect types.

[0018] Furthermore, a phase unwrapping step is included, in which the continuous true phase Φ(x,y) is restored by a phase unwrapping algorithm, that is, Φ(x,y)=Unwrap(Φw(x,y)).

[0019] Furthermore, in the reflection measurement mode, the depth map h(x,y) of the object and the phase distortion introduced by it is proportional to , where K is the scale factor obtained through system calibration and is related to the system's geometry. The depth map h(x,y) corresponds to the depth map value in the reflectance measurement. The scale factor K quantifies the phase distortion into actual height, making the depth map's physical meaning clear and enabling direct measurement of defect dimensions (such as bump height and depression depth). The depth map h(x,y) refers to the continuous true phase obtained through phase unwrapping in reflective phase deflectometry, combined with the scale factor K obtained through system calibration. This image data reflects the three-dimensional topography of the measured product surface, and its value directly corresponds to the bump height or depression depth of the measured surface.

[0020] Furthermore, the preset logic for defect identification and classification includes: Concave and convex point determination: Concave and convex points appear as local depressions or protrusions on the depth map obtained by reflectance measurement, but do not show significant changes on the reflectance modulation map or the backlight modulation map. Foreign matter identification: Surface foreign matter appears as a bump on the depth map of reflectance measurement and a noticeable dark area on the corresponding reflectance modulation map. Interlayer foreign matter does not show any abnormality on the depth map of reflectance measurement, but causes significant disturbance on the phase difference map of backlight transmission measurement and a noticeable dark area on the backlight modulation map. Backlight transmission measurement involves projecting multi-frequency phase-shifted fringe light through the product under test, capturing the transmitted fringe image with a camera, and performing zero-reference plane differential measurement. Scratches appear on the depth map or phase difference map. Scratch detection: Defective areas appear as high-aspect-ratio stripes on depth or phase difference images. By matching multi-dimensional data (depth, phase difference, and modulation), the system accurately classifies concave and convex spots, surface foreign matter, interlayer foreign matter, and scratches. Preset logic and quantitative criteria eliminate ambiguity in manual judgment and ensure consistent test results.

[0021] The nonlinear screen appearance defect evaluation method of the present invention has the following technical effects: High-precision detection: By measuring phase rather than light intensity, this system is extremely sensitive to minute light deflections caused by defects. A 65-megapixel industrial camera can capture subtle image changes. Combined with multi-frequency phase-shifted stripe light projection and precise phase calculation, the system can detect tiny defects that are difficult to detect with traditional methods, significantly improving detection accuracy.

[0022] Eliminate systematic errors: The zero reference plane measurement method collects a reference phase image without a sample and a measured phase image with a sample and subtracts them, effectively eliminating inherent system errors such as camera lens distortion and uneven light source screen. This eliminates the need for complex and expensive calibration processes, reduces system costs, and improves the reliability of measurement results.

[0023] Accurate defect classification: By combining depth maps, phase difference maps, and reflection / backlight dual modulation maps, supplemented by geometric parameter analysis, a reliable defect classification logic has been established. This system can accurately distinguish between different types of defects, including concave and convex spots, surface foreign matter, interlayer foreign matter, and scratches, resolving the issue of confusion in defect attribution in existing technologies.

[0024] Complete, dual-sided inspection: By flipping the product and measuring both sides, we ensure equally high-precision inspection of both surfaces. Reflective measurement targets surface defects, while transmissive measurement reveals internal and complex flaws. This combination enables the system to comprehensively detect all defects on transparent parts, eliminating the potential omissions that can occur with single-sided inspection.

[0025] In order to more clearly illustrate the structural features and effects of the present invention, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 Schematic diagram of hardware functional modules of a transparent component defect detection system based on dual-optical path phase deflectometry of the present invention; Figure 2 The present invention is a flow chart of a transparent component defect detection method based on dual-light path phase deflectometry.

[0027] Figure 3 The diagram is a defect classification logic diagram of a transparent component defect detection method based on dual-light-path phase deflectometry according to the present invention. DETAILED DESCRIPTION

[0028] The present invention will be further described below with reference to the accompanying drawings and related knowledge, and described clearly and completely. Obviously, the described applications are only part of the embodiments of the present invention, rather than all of the embodiments.

[0029] Reference Figure 1 As shown, embodiment 1: the present invention provides a transparent part defect detection system based on dual-light path phase deflectometry, including a camera 1, a reflective light source 2, a spectroscope 3, a backlight source 4 and a processing unit; Camera 1: A 65-megapixel industrial camera placed directly above the product under test 5, used to capture fringe images; Reflective light source 2: located on the side of the system, used to project multi-frequency phase-shifted stripe light.

[0030] Beam splitter 3: It is set at a specific angle between the camera and the product under test, and can reflect the stripe light projected by the reflected light source to the upper surface of the product under test.

[0031] Backlight source 4: It is located directly below the product under test 5 and is a programmable screen (such as LCD or OLED) used to project multi-frequency phase-shifted stripe light through the product under test.

[0032] Processing unit: electrically connected to the camera, reflective light source and backlight source, used to control the projection timing of the reflective light source and backlight source, process the images collected by the camera and execute the defect detection algorithm.

[0033] Example 2, reference Figure 2 、 Figure 3 As shown, a transparent component defect detection method based on dual-path phase deflectometry includes the following core steps: First measurement (A side 6 detection): Specifically, for reflectance measurement of surface A 6, the reflective light source is activated to perform phase deflectometry on surface A 6 (the top surface of product 5). Based on the captured image, the processing unit calculates a depth map reflecting the three-dimensional topography of surface A and a modulation map reflecting the fringe contrast.

[0034] Overall transmission measurement: Start the backlight source and perform zero reference plane differential measurement. First, collect the reference phase image (Φ ref ), and then collect the measured phase diagram with samples (Φ test ), and obtain the phase difference diagram At the same time, the processing unit also calculates the backlight modulation map based on the transmission fringe image.

[0035] Product Flip: In order to ensure that defects on the upper and lower surfaces are not missed, after the first measurement is completed, the product under test is turned 180 degrees by a robot or manually so that its B surface 7 (the lower surface of the product under test 5) faces the camera.

[0036] Measure again (B side 7 detection): B-side reflection measurement: Repeat the steps for the A-side reflection measurement and perform reflective phase deflectometry on the B-side of the product to obtain the depth map and modulation map of the B-side.

[0037] Comprehensive defect analysis and classification: The processing unit integrates all the above measurement data and identifies and classifies defects according to the following logic: Concave and convex point determination: This appears as a local depression or convexity on the depth map obtained by reflection measurement, but usually has no significant changes on the reflection and backlight modulation maps.

[0038] Foreign matter determination: Surface foreign matter: appears as a bump on the depth map of the reflectance measurement, and as a noticeable dark area (reduced modulation) on the corresponding reflectance modulation map.

[0039] Interlayer foreign matter: There is no abnormality in the depth map of the reflection measurement, but there is an abnormality in the phase difference map of the back-transmission measurement. There is significant disturbance on the , and obvious dark areas appear on the backlight modulation graph.

[0040] Scratch determination: On the depth map or phase difference map, the defect area appears as a stripe feature with a high aspect ratio.

[0041] This invention enables: high-precision detection, by measuring phase rather than light intensity, making it extremely sensitive to minute light deflections caused by defects. It also eliminates system errors, effectively eliminating inherent system errors through a zero-reference plane measurement method. It also enables precise defect classification, combining depth maps, phase difference maps, and reflection / backlight dual modulation maps, supplemented by geometric parameter analysis, to establish a reliable defect classification logic. It also ensures complete double-sided inspection without omissions: By flipping the product and performing bilateral measurement, it ensures equally high-precision inspection of both the front and back surfaces of the product.

[0042] Example 2, a transparent component defect detection method based on dual-path phase deflectometry, System construction: Figure 1 As shown in Figure 1, the system hardware includes a 65-megapixel industrial camera, a spectroscope, a sample stage, a back-transmitted LCD screen, and a side-reflective light source. All components are connected to an industrial computer (processing unit).

[0043] The core of the transparent part defect detection method of the present invention lies in accurately extracting information such as phase and modulation from multiple collected fringe images.

[0044] The N-step phase shift method extracts the gray value I of the i-th fringe image captured by the camera. i (x,y) can be expressed as: I i (x,y)=A(x,y)+B(x,y)cos[Φ(x,y)+δ i ]; Among them, (x, y) is the pixel coordinate, A(x, y) is the background light intensity, B(x, y) is the fringe amplitude, Φ(x, y) is the phase to be determined, δ i =2πi / N is the phase shift of the i-th step, and N is the total number of steps (N is usually 4 in this embodiment).

[0045] By solving the above equations, we can obtain the wrapped phase diagram Φw(x,y) (its range is (-π,π): ; For the commonly used four-step phase shift (δ i =0,π / 2,π,3π / 2), the above formula is simplified to: ; At the same time, the fringe amplitude map B(x,y) and the background light intensity map A(x,y) can be calculated: ; ; The modulation map M(x,y) is defined as the ratio of the amplitude to the background intensity, reflecting the clarity of the fringe: ; Foreign matter or partial surface coating defects can cause local light absorption or scattering, resulting in a sharp drop in B(x,y), which appears as a dark area in the modulation map M(x,y). This is the key basis for determining such defects.

[0046] Phase unwrapping: The Φw(x,y) obtained by the above method is wrapped in the interval (-π,π). A phase unwrapping algorithm (such as the multi-frequency heterodyne method or the Gray code-assisted method) is needed to restore its continuous true phase Φ(x,y).

[0047] Φ(x,y)=Unwrap(Φw(x,y)); The zero reference plane difference method is the core of the present invention in transmission measurement. By subtracting two measurements, the system error is eliminated.

[0048] ; where Φ ref is the system phase measured at no load, Φ test is the phase measured after loading the sample. It accurately reflects the phase distortion caused only by the sample (including its internal and external defects).

[0049] Depth map reconstruction In reflection measurement mode, the depth map h(x,y) of the object and the phase distortion introduced by it (relative to an ideal plane).

[0050] Where K is the scaling factor obtained through system calibration and is related to the system geometry (such as the angle between the camera and the light source, the working distance, etc.). This depth map h(x,y) is directly used to identify 3D topographic defects such as dents.

[0051] Detection process example: Phase 1: Surface A detection, Surface A reflection measurement: The computer controls the reflected light source to project four-step phase-shifted fringes, and the camera collects four images I1, I2, I3, and I4. The processing unit calculates the depth map h of Surface A according to the above formula. A (x,y) and modulation map M A (x,y). Overall transmission measurement: Computer controls the backlight source, first removes the glass, collects reference fringes and calculates the continuous reference phase map Φ ref (x,y). Then put the glass back, collect the sample fringes and calculate the measured phase map Φ test (x,y). Finally, the phase difference map is obtained by subtraction and backlight modulation diagram M BG (x,y).

[0052] The second stage: The robotic arm in the product flip detection system flips the glass to be tested 180 degrees so that side B faces upward.

[0053] Phase 3: B-side reflection measurement: Repeat the A-side reflection measurement and perform reflection measurement on the B-side to obtain the depth map h of the B-side. B (x,y) and modulation map M B (x,y).

[0054] The fourth stage: comprehensive analysis and processing of defects Analyze multi-dimensional data. Calculate the gradient of the phase difference image or depth image to enhance the defect edge signal: ; By performing threshold segmentation on the gradient map G(x,y), the initial outline of the defect is determined. Then, the feature values ​​of each data map within the outline are combined and classified according to the preset logic.

[0055] It should be noted that in the present invention, when there is no transparent part to be measured in the optical path, the reference stripes are projected by the backlight source and captured by the camera to calculate the reference phase map (Φ ref ); Place the transparent part to be measured in the light path, project the measurement fringes through the backlight source, and collect them by the camera to calculate the measurement phase diagram (Φ test ) ; Subtracting the measured phase image from the reference phase image to obtain a phase difference image reflecting the defects of the transparent element itself , its mathematical expression is: .

[0056] In the present invention, it is necessary to perform at least one reflective phase deflectometry measurement on a transparent member to obtain a three-dimensional depth map h(x,y) of its surface. After completing the initial measurement, the transparent member is flipped over. The reflective phase deflectometry measurement is repeated on the other surface of the flipped transparent member. Furthermore, in the defect classification step, the classification step includes: determining a concave-convex point (Dent) defect based on a local height change in the depth map h(x,y); In a preferred embodiment of the present invention, a defect classification step is also included, which includes: calculating the reflection modulation map and the backlight modulation map M(x, y) from the collected reflection fringe image and transmission fringe image respectively; determining the surface foreign matter defect based on the raised features in the depth map and the modulation reduction in the reflection modulation map; determining the interlayer foreign matter defect based on the phase disturbance in the phase difference map and the modulation reduction in the backlight modulation map, or performing geometric parameter analysis on the defect area identified in the depth map or the phase difference map, and determining the defect with an aspect ratio greater than a preset threshold as a scratch defect.

[0057] This invention utilizes the core logic of dual-optical path measurement, product flipping, and multi-dimensional data fusion analysis. By combining reflected and transmitted optical paths with precise phase calculation and defect classification algorithms, it achieves comprehensive detection of internal and external defects in transparent parts. The following is a detailed description of the specific working steps: After system initialization and parameter configuration, hardware connection and calibration, the processing unit first completes connectivity checks for all hardware modules, including communication tests for the 65-megapixel industrial camera, the side-mounted reflective light source, the beam splitter (tilted 45° between the camera and the sample stage), and the backlight source (LCD / OLED screen, located directly below the sample stage). Using pre-stored calibration parameters, the processing unit sets system geometric parameters, such as the angle between the camera and the light source and the working distance. It also determines the scaling factor K used in depth map calculations to ensure accurate conversion between phase and actual height.

[0058] Light source parameter configuration: the processing unit sends control instructions to the reflective light source and backlight source to set the parameters of the multi-frequency phase-shifted stripes: The number of phase shift steps N=4; the fringe frequency is adjusted according to the thickness of the transparent part being measured.

[0059] First measurement (A-side detection stage): Step 1: A-surface reflection measurement (detection of upper surface defects); The optical path is started, the processing unit controls the reflected light source to turn on, and the stripe light is reflected by the beam splitter and vertically illuminates the A surface (upper surface) of the transparent part to be measured. The reflected light is transmitted through the beam splitter and enters the camera lens.

[0060] Image acquisition,The camera continuously collects four fringe images according to the phase shift sequence,,and each image corresponds to a different phase shift amount.

[0061] Data Processing: Wrapped Phase Calculation: The processing unit calculates the wrapped phase map using a four-step phase shift formula. Phase Unwrapping: A multi-frequency heterodyne method is used to eliminate phase jumps and restore the continuous true phase. Depth Map Generation: Based on the proportional relationship between the depth map and phase distortion, a 3D depth map of the A-surface (reflecting convexity and concavity) is generated. Reflection Modulation Map Calculation: A modulation map is generated by calculating the ratio of amplitude to background light intensity (reflecting fringe clarity; foreign matter can cause localized dark areas).

[0062] Step 2: Overall transmission measurement (detection of internal and double-sided comprehensive defects); Reference phase acquisition (zero reference plane calibration): The processing unit controls the robot to remove the test piece on the sample stage. The backlight source is turned on and projects four-step phase-shifted stripes. The camera captures the reference stripe image without a sample and calculates the system's inherent phase (including system errors such as lens distortion and screen unevenness).

[0063] For phase measurement, the robot places the test piece back on the sample stage (side A facing up). The backlit light source projects the same stripes again. The light passes through the transparent piece and is captured by the camera. The measurement phase including the sample defects is calculated.

[0064] Phase difference and backlight modulation calculation, phase difference map generation: Eliminate systematic errors through differentiation: This map only reflects the light deflection caused by defects inside the sample (such as interlayer foreign matter) and double-sided defects.

[0065] Backlight modulation map calculation: Use the same method as reflection measurement.

[0066] The product is flipped (enabling comprehensive double-sided inspection). The processing unit controls the robotic arm to grip the transparent part under test and rotate it 180° along its vertical axis, positioning the original bottom surface (Side B) upward. This ensures that Side B and Side A are aligned in the same optical path during measurement, minimizing inspection errors caused by geometric deviations. After flipping, the sample stage is repositioned to ensure that the center of the part under test is aligned with the camera's optical axis.

[0067] Measure again (B side detection stage), repeat all the steps of "A side reflection measurement" and detect the flipped B side: The reflected light source projects stripes, and the camera captures four stripe images reflected from surface B. The processing unit calculates the depth map and reflection modulation map of surface B. The data is stored in the buffer area of ​​the processing unit and associated with the surface A and transmission measurement data.

[0068] Comprehensive defect analysis and classification, processing unit integration Measurement data is used to identify defects through the following logic: Concave and convex point determination, h A and h B Perform local gradient analysis. If there are sudden changes in the height of consecutive pixels (such as depressions / convexities), and the M at the corresponding position A 、M B 、M BG There is no significant change and it is judged as a concave or convex point.

[0069] Surface foreign matter determination, if h A (or h B ) has an isolated convex region, and this region is in M A (or M B ) is manifested as a significant decrease in modulation, which is judged to be a foreign object on the surface of surface A (or surface B) (the foreign object obstructs the stripes, resulting in reduced stripe clarity).

[0070] Determination of interlayer foreign matter, if h A and h B No abnormalities, but There is a local phase disturbance in the corresponding position M BG The modulation index decreased, which was determined to be an internal foreign object (the foreign object scattered the light, causing phase distortion and a decrease in modulation index).

[0071] Scratch determination: perform edge detection on the depth map or phase difference map and calculate the aspect ratio of the defect area: if the length / width is greater than 5 (preset threshold) and is distributed in a continuous strip, it is determined to be a scratch (surface scratches are within h A / h B In the middle of the development, the internal scratches mid-development).

[0072] Result output and feedback: The processing unit generates a test report based on the test results (defect type, location coordinates, size parameters), displays it in real time on the display screen, and triggers the sorting signal: In the present invention, the reflected light path focuses on the surface three-dimensional morphology (depth map) and surface occlusion (reflection modulation), and the transmitted light path focuses on internal defects (phase difference map) and internal occlusion (backlight modulation). The data of the two complement each other to achieve full coverage of surface and internal defects.

[0073] Through the fusion analysis of multi-dimensional data (depth, phase difference, modulation depth), the limitations of single-parameter detection have been broken through.

[0074] The technical principles of the present invention have been described above in conjunction with specific embodiments, which are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention fall within the scope of protection of the present invention. Those skilled in the art will be able to conceive of other specific embodiments of the present invention without inventive effort, and such methods will fall within the scope of protection of the present invention.

Claims

1. A transparent component defect detection system based on dual-path phase deflectometry, characterized in that: It includes a camera, a reflected light source, a spectrometer, a backlight source and a processing unit; the camera is arranged directly above the product under test and is used to collect stripe images; the reflected light source is used to project multi-frequency phase-shifted stripe light; the spectrometer is arranged between the camera and the product under test and can reflect the stripe light projected by the reflected light source to the upper surface of the product under test; the backlight source is arranged directly below the product under test and is a programmable screen used to project multi-frequency phase-shifted stripe light through the product under test; the processing unit is electrically connected to the camera, the reflected light source and the backlight source, and is used to control the projection timing of the reflected light source and the backlight source, process the images collected by the camera and execute the defect detection algorithm.

2. The system according to claim 1, wherein: The camera is an industrial camera with a resolution of 50-80 million pixels.

3. The system according to claim 1, wherein: The backlight source is an LCD screen or an OLED screen.

4. A method for detecting defects in transparent parts based on dual-path phase deflectometry, the method being used in the transparent part defect detection system based on dual-path phase deflectometry according to any one of claims 1 to 3, characterized in that: The following steps are involved: Step S1: Start the reflective light source and perform reflective phase deflectometry on surface A of the product under test. The camera collects the reflected fringe image. The processing unit calculates a depth map reflecting the three-dimensional topography of surface A and a reflective modulation map reflecting the fringe contrast based on the collected reflected fringe image. Simultaneously, start the backlight source and project multi-frequency phase-shifted fringe light through the product under test. The camera collects the transmitted fringe image and performs zero-reference plane differential measurement. That is, first collects a reference phase map without a sample, then collects a measured phase map with a sample to obtain a phase difference map. At the same time, the processing unit calculates the backlight modulation map based on the transmitted fringe image. Step S2: After the first measurement is completed, the product to be measured is turned over so that its B side faces the camera; Step S3: Repeat the reflective phase deflectometry measurement steps for surface A in step S1 to perform reflective phase deflectometry on surface B of the product. The camera captures a reflection fringe image, and the processing unit obtains a depth map and a reflection modulation map of surface B based on the captured reflection fringe image. Step S4: The processing unit integrates all the above measurement data and identifies and classifies defects according to preset logic.

5. The method according to claim 4, characterized in that In the reflective phase deflectometry measurement of surface A and surface B, the N-step phase shift method is used to extract the package phase and modulation. The gray value I of the i-th fringe image captured by the camera is i (x,y) is represented as: I i (x,y)=A(x,y)+B(x,y)cos[Φ(x,y)+δ i ], where (x, y) is the pixel coordinate, A(x, y) is the background intensity map, B(x, y) is the fringe amplitude map, Φ(x, y) is the phase to be determined, δ i =2πi / N is the phase shift of the i-th step, and N is the total number of steps.

6. The method according to claim 5, characterized in that N is 4, four-step phase shift δ i =0,π / 2,π,3π / 2, the wrapped phase diagram Φw(x,y) is simplified to: .

7. The method according to claim 5, characterized in that The calculation formulas for the fringe amplitude map B(x,y) and the background light intensity map A(x,y) are: , ; The modulation map M(x,y) is defined as the ratio of the fringe amplitude map to the background intensity map, that is, .

8. The method according to claim 6, characterized in that The process also includes a phase unwrapping step, in which the continuous true phase Φ(x,y) is restored through a phase unwrapping algorithm, that is, Φ(x,y)=Unwrap(Φw(x,y)).

9. The method according to claim 4, characterized in that In reflection measurement mode, the depth map h(x,y) of the object and the phase distortion introduced by it is proportional to , where K is the scaling factor obtained through system calibration, and the depth map h (x, y) corresponds to the value of the depth map in the reflection measurement.

10. The method according to claim 4, characterized in that The preset logic for defect identification and classification includes: bumps and concave spots appear as local depressions or protrusions on the depth map obtained by reflection measurement, but there is no significant change on the reflection modulation map and the backlight modulation map; surface foreign matter appears as a protrusion on the depth map of reflection measurement, and at the same time appears as a significant dark area on its corresponding reflection modulation map; interlayer foreign matter has no abnormality on the depth map of reflection measurement, but has significant disturbances on the phase difference map of backlight measurement, and appears as a significant dark area on the backlight modulation map; backlight measurement refers to projecting multi-frequency phase-shifted stripe light through a backlight light source to penetrate the product under test, the camera captures the transmitted stripe image and performs zero reference plane differential measurement; scratches appear as strip-shaped features with a high aspect ratio on the depth map or phase difference map.

Images