An automatic board testing machine

Abstract

This application relates to the field of automated testing equipment technology, and discloses an automatic LED board splitting and testing machine, including a worktable. A feeding component is provided on one side of the upper surface of the worktable for conveying unsegmented LED boards. A splitting component is provided in the middle of the upper surface of the worktable for splitting large unsegmented LED boards into smaller pieces. A recycling component is provided on the other side of the upper surface of the worktable for collecting the split and tested LED boards. A transfer component is provided on the upper surface of the worktable for transporting the LED boards to a specific location. The transfer component is located at the output end of the feeding component and the splitting component. A testing component is provided between the splitting component and the recycling component for detecting whether the LED boards are usable. By setting up the testing component and the secondary splitting mechanism, parallel processing of testing and splitting is achieved, shortening the processing and testing cycle of a single LED board.

Description

Technical Field

[0001] This invention relates to the field of automated testing equipment technology, specifically an automatic board separation testing machine. Background Technology

[0002] With the widespread application of LED lighting technology and the increasing market demand, the production efficiency and quality control of LED light boards have become particularly critical. The manufacturing process of LED light boards typically involves dividing a large PCB board into smaller, independent units, followed by electrical performance testing and appearance defect detection of these units. However, existing automated production lines face numerous challenges in performing these steps.

[0003] On the one hand, traditional processes often treat board separation and testing as independent, sequential steps. That is, a PCB board must be completely separated before it can be sent to the testing station for inspection. This sequential operation mode leads to a longer production cycle, limiting overall capacity and failing to meet the demands of efficient, high-volume production. Especially in the face of increasingly shorter production cycles, optimizing the production process to achieve a more compact and faster operating mode is a pressing issue for the industry. On the other hand, in the quality inspection of LED light boards, especially the detection of appearance defects, traditional machine vision systems are widely relied upon. These systems typically acquire and analyze images under a fixed light source and a single viewing angle. However, LED light board surfaces often exhibit varying degrees of specular reflection, texture differences, and subtle morphological variations due to production tolerances or material properties. Under fixed lighting conditions, these factors easily lead to problems such as high gloss saturation, shadow occlusion, or insufficient contrast in the images, making it difficult for vision inspection systems to accurately capture minute defects such as scratches, foreign objects, and solder joint defects. When the ambient light fluctuates or different batches and models of LED light boards need to be tested, a single lighting strategy is often difficult to adapt, leading to an increase in the false positive rate or the missed detection rate, which seriously affects the reliability of the test and the effective control of product quality.

[0004] Therefore, developing an intelligent detection method that can adapt to environmental changes, overcome surface optical interference, and improve the accuracy of defect detection has become a key technological bottleneck in improving the production quality of LED light boards. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an automatic board separation and testing machine, which solves the problems of low efficiency and insufficient accuracy in electrical and appearance testing of existing LED light boards after separation, as well as the susceptibility of single testing mode to ambient light and surface characteristics.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] An automatic LED board splitting and testing machine includes a workbench. A feeding component is located on one side of the upper surface of the workbench for conveying unsegmented LED boards. A splitting component is located in the middle of the upper surface of the workbench for splitting large unsegmented LED boards into smaller pieces. A recycling component is located on the other side of the upper surface of the workbench for collecting the split and tested LED boards. A transfer component is located on the upper surface of the workbench for transporting the LED boards to a specific location. The transfer component is located at the output end of the feeding component and the splitting component. A testing component is located between the splitting component and the recycling component for detecting whether the LED boards can be used normally. The testing component integrates a detection system.

[0008] The dividing assembly includes two support columns, which are fixedly connected to the upper surface of the workbench. A motor is fixedly connected to the outer surface of one of the support columns. A placement plate is fixedly connected to the output end of the motor. The placement plate is rotatably connected between the two support columns. A cylinder is fixedly connected to the top of the placement plate. A dividing blade is fixedly connected to the output end of the cylinder. The dividing blade is slidably connected inside the placement plate.

[0009] The segmentation component also includes an auxiliary component. The output end of the auxiliary component is fixedly connected to an impact arc, which is rotatably connected to the outside of the auxiliary component. A first segmentation strip is fixedly connected to the top end of the auxiliary component, and a second segmentation strip is rotatably connected to one end of the first segmentation strip.

[0010] Preferably, the auxiliary component includes a cylinder three, which is fixedly connected to the upper surface of the workbench and located below the placement plate. An auxiliary claw is fixedly connected to the output end of the cylinder three. A column is fixedly connected to the upper surface of the workbench. A motor two is fixedly connected to the side surface of the column. An impact arc is fixedly connected to the output end of the motor two. The impact arc is rotatably connected to the outside of the column. A dividing strip one is fixedly connected to the top of the column.

[0011] Preferably, the feeding assembly includes a telescopic frame and a support frame. The telescopic frame and the support frame are fixedly connected to one side of the upper surface of the workbench. Two cylinders are fixedly connected to one side of the upper surface of the workbench. A linkage plate is fixedly connected to the output end of each cylinder. On both sides of the cylinders, multiple limiters are fixedly connected to one side of the upper surface of the workbench. A limit rod is slidably connected inside each limiter. One end of the limit rod is fixedly connected to the outer surface of the linkage plate. Below the support frame, an electric lead screw is fixedly connected to the upper surface of the workbench. A conveying block is threadedly connected to the outer surface of the electric lead screw.

[0012] Preferably, the transfer assembly includes a gantry frame, which is fixedly connected to the upper surface of the workbench. Two grippers are slidably connected to the outer surface of the gantry frame. A vertical plate is fixedly connected to the upper surface of the workbench below the gantry frame. An electric slide rail is fixedly connected to one side of the outer surface of the vertical plate. A rotating block is slidably connected to the outer surface of the electric slide rail.

[0013] Preferably, the testing assembly includes two cylinders, which are fixedly connected to the upper surface of the worktable and located on both sides of the dividing strip. Detection electrodes are fixedly connected to the output ends of the cylinders. A main camera and an auxiliary camera are fixedly connected to the upper surface of the worktable and are located above the dividing strip. A rotating arm is fixedly connected to the upper surface of the worktable, and a supplementary light is fixedly connected to the output end of the rotating arm. The testing assembly also includes a processor, which is fixedly connected to the top of the gantry frame, and the detection system is integrated inside the processor.

[0014] Preferably, the recycling assembly includes a telescopic frame 2 and a support frame 2. The telescopic frame 2 and the support frame 2 are fixedly connected to the other side of the upper surface of the workbench. Two cylinders 5 are fixedly connected to the other side of the upper surface of the workbench. A linkage plate 2 is fixedly connected to the output end of the cylinders 5. On both sides of the cylinders 5, a plurality of limiters 2 are fixedly connected to one side of the upper surface of the workbench. Limiting rods 2 are slidably connected inside the limiters 2. One end of the limiting rods 2 is fixedly connected to the outer surface of the linkage plate 2. Below the support frame 2, an electric lead screw 2 is fixedly connected to the upper surface of the workbench. A conveying block 2 is threadedly connected to the outer surface of the electric lead screw 2.

[0015] Preferably, the detection system includes:

[0016] The image acquisition module is used to acquire images of the LED light panel;

[0017] A light source module is used to provide illumination for the image acquisition module;

[0018] The processing control module includes:

[0019] An uncertainty analysis unit is used to receive the first image acquired by the image acquisition module and calculate the confidence score and pixel-level uncertainty spectrum of the area to be inspected of the LED light board.

[0020] An active control decision unit is used to select a second set of illumination parameters from a predefined strategy library based on the characteristics of the uncertainty spectrum when the confidence score is lower than a preset threshold or the uncertainty spectrum indicates the existence of an uncertainty region, and instruct the supplementary light to execute the strategy.

[0021] The multimodal fusion decision unit is used to fuse information from multiple images acquired under different illumination parameters to obtain the final detection result of the LED light panel.

[0022] Preferably, the predefined strategy library of the active control decision-making unit includes:

[0023] The first control strategy is to activate the LED beads with low illumination angle in the fill light when the uncertainty area exhibits high light saturation characteristics, and control the rotating arm to rotate the fill light by a certain angle.

[0024] The second control strategy involves controlling the supplementary light to switch between two different LED beads with different illumination angles in a short period of time when the uncertain area exhibits low contrast characteristics.

[0025] Preferably, the step of the multimodal fusion decision unit obtaining the final detection result of the LED light panel specifically includes:

[0026] Based on the image pairs synchronously acquired by the main camera and the auxiliary camera, the relative height map of the area to be inspected of the LED light panel is calculated using a stereo vision algorithm;

[0027] Images acquired under different lighting parameters and the relative height map are used as multiple evidence sources. Independent decisions are made on the same area to be inspected, and multiple local decision results are obtained. The confidence scores corresponding to each local decision result are then weighted, and the final detection result is output using a weighted voting method.

[0028] Preferably, the processing control module further includes a strategy library update unit, which stores the illumination parameter strategies and corresponding uncertainty characteristics successfully executed by the active control decision unit for a specific type of LED light panel into the predefined strategy library.

[0029] This invention provides an automatic board separation and testing machine. It has the following beneficial effects:

[0030] 1. This invention achieves a parallel processing flow by setting up a test component and a secondary segmentation mechanism (which works in concert with motor two and impact arc) so that while one half of the LED light board is being tested for electrical and visual properties on segmentation strip one, the other half of the LED light board is being impacted and segmented on segmentation strip two. This design, which parallelizes two time-consuming steps that originally needed to be executed serially, shortens the overall processing and testing cycle of a single light board.

[0031] 2. This invention, through the uncertainty analysis unit inside the detection system, can first quantitatively evaluate the quality of the initial image. When factors causing uncertainty, such as highlights and shadows, are detected, the active control decision unit will instruct the programmable multi-angle supplementary light to change the illumination parameters to obtain a higher quality image. Finally, the multimodal fusion decision unit fuses and decides the two-dimensional images under different illuminations with the three-dimensional relative height map information generated by the dual-camera system, effectively suppressing false detections and false negatives under single imaging conditions.

[0032] 3. By setting up a strategy library update unit, this invention enables the association and storage of image features that lead to uncertainty with the lighting parameter strategy that successfully solves the problem after a successful dynamic lighting adjustment. This allows the system to accumulate experience during use and autonomously optimize its detection strategy for the specific reflectivity or surface characteristics of different products, reducing the need for readjustment when changing production lines or ambient light. Attached Figure Description

[0033] Figure 1 This is a perspective view of the device of the present invention;

[0034] Figure 2 This is a side view of the device of the present invention;

[0035] Figure 3 This is a schematic diagram of the execution component of the present invention;

[0036] Figure 4 This is a schematic diagram of the feeding component of the present invention;

[0037] Figure 5 This is a partition diagram of the working components of the present invention;

[0038] Figure 6 This is a schematic diagram of the test components of the present invention;

[0039] Figure 7 This is a schematic diagram of the segmentation component of the present invention;

[0040] Figure 8 This is a schematic diagram of the recycling component of the present invention;

[0041] Figure 9 This is a schematic diagram of the transfer component of the present invention;

[0042] Figure 10 This is a schematic diagram of the supplementary lighting of the present invention;

[0043] Figure 11 This is a flowchart of the processing logic of the detection system of the present invention;

[0044] Figure 12 This is a schematic diagram illustrating an example of the active optical control strategy of the present invention.

[0045] The components include: 1. Workbench; 2. Feeding assembly; 3. Segmentation assembly; 4. Recycling assembly; 5. Transfer assembly; 6. Testing assembly; 7. Detection system; 301. Support column; 302. Motor 1; 303. Placement plate; 304. Cylinder 1; 305. Segmentation blade; 201. Telescopic frame 1; 202. Support frame 1; 203. Cylinder 2; 204. Linkage plate 1; 205. Limiter 1; 206. Limiting rod 1; 207. Electric lead screw 1; 208. Conveying block 1; 501. Gantry frame; 502. Gripper; 503. Vertical plate; 504. Electric slide rail. 505. Rotating block; 306. Cylinder 3; 307. Auxiliary claw; 308. Column; 309. Motor 2; 310. Impact arc; 311. Dividing strip 1; 312. Dividing strip 2; 601. Cylinder 4; 602. Detection electrode; 603. Main camera; 604. Auxiliary camera; 605. Rotating arm; 606. Fill light; 607. Processor; 401. Telescopic frame 2; 402. Support frame 2; 403. Cylinder 5; 404. Linkage plate 2; 405. Limiter 2; 406. Limiting rod 2; 407. Electric lead screw 2; 408. Conveying block 2. Detailed Implementation

[0046] 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 some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] Please see the appendix Figure 1 - Appendix Figure 12 This invention provides an automatic LED board splitting and testing machine, including a workbench 1. A feeding component 2 is provided on one side of the upper surface of the workbench 1 for conveying unsegmented LED boards. A splitting component 3 is provided in the middle of the upper surface of the workbench 1 for splitting large unsegmented LED boards into smaller pieces. A recycling component 4 is provided on the other side of the upper surface of the workbench 1 for collecting the split and tested LED boards. A transfer component 5 is provided on the upper surface of the workbench 1 for transporting the LED boards to a specific position. The transfer component 5 is located at the output end of the feeding component 2 and the splitting component 3. A testing component 6 is provided between the splitting component 3 and the recycling component 4 for detecting whether the LED boards can be used normally. The testing component 6 integrates a detection system 7.

[0048] The dividing component 3 includes two support columns 301. The support columns 301 are fixedly connected to the upper surface of the workbench 1. A motor 302 is fixedly connected to the outer surface of one of the support columns 301. A placement plate 303 is fixedly connected to the output end of the motor 302. The placement plate 303 is rotatably connected between the two support columns 301. A cylinder 304 is fixedly connected to the top of the placement plate 303. A dividing blade 305 is fixedly connected to the output end of the cylinder 304. The dividing blade 305 is slidably connected inside the placement plate 303.

[0049] The feeding assembly 2 includes a telescopic frame 201 and a support frame 202. The telescopic frame 201 and the support frame 202 are fixedly connected to one side of the upper surface of the workbench 1. Two cylinders 203 are fixedly connected to one side of the upper surface of the workbench 1. A linkage plate 204 is fixedly connected to the output end of the cylinders 203. On both sides of the cylinders 203, multiple limiters 205 are fixedly connected to one side of the upper surface of the workbench 1. Limiting rods 206 are slidably connected inside the limiters 205. One end of the limiting rods 206 is fixedly connected to the outer surface of the linkage plate 204. Below the support frame 202, an electric lead screw 207 is fixedly connected to the upper surface of the workbench 1. A conveying block 208 is threadedly connected to the outer surface of the electric lead screw 207.

[0050] The transfer assembly 5 includes a gantry frame 501, which is fixedly connected to the upper surface of the workbench 1. Two grippers 502 are slidably connected to the outer surface of the gantry frame 501. Below the gantry frame 501, a vertical plate 503 is fixedly connected to the upper surface of the workbench 1. An electric slide rail 504 is fixedly connected to one side of the outer surface of the vertical plate 503. A rotating block 505 is slidably connected to the outer surface of the electric slide rail 504.

[0051] The dividing assembly 3 also includes a cylinder 306, which is fixedly connected to the upper surface of the worktable 1 and located below the placement plate 303. An auxiliary claw 307 is fixedly connected to the output end of the cylinder 306. A column 308 is fixedly connected to the upper surface of the worktable 1. A motor 309 is fixedly connected to the side surface of the column 308. An impact arc 310 is fixedly connected to the output end of the motor 309. The impact arc 310 is rotatably connected to the outside of the column 308. A dividing strip 311 is fixedly connected to the top of the column 308. A dividing strip 312 is rotatably connected to one end of the dividing strip 311.

[0052] Test component 6 includes two cylinders 601, which are fixedly connected to the upper surface of workbench 1. The two cylinders 601 are located on both sides of the dividing strip 311. The output end of cylinder 601 is fixedly connected to a detection electrode 602. A main camera 603 and an auxiliary camera 604 are fixedly connected to the upper surface of workbench 1. The main camera 603 and the auxiliary camera 604 are located above the dividing strip 311. A rotating arm 605 is fixedly connected to the upper surface of workbench 1. A supplementary light 606 is fixedly connected to the output end of the rotating arm 605. Test component 6 also includes a processor 607, which is fixedly connected to the top of gantry 501. The processor 607 integrates a detection system 7.

[0053] The recycling component 4 includes a telescopic frame 401 and a support frame 402. The telescopic frame 401 and the support frame 402 are fixedly connected to the other side of the upper surface of the workbench 1. Two cylinders 403 are fixedly connected to the other side of the upper surface of the workbench 1. The output end of the cylinders 403 is fixedly connected to a linkage plate 404. On both sides of the cylinders 403, multiple limiters 405 are fixedly connected to one side of the upper surface of the workbench 1. Limiting rods 406 are slidably connected inside the limiters 405. One end of the limiting rods 406 is fixedly connected to the outer surface of the linkage plate 404. Below the support frame 402, an electric lead screw 407 is fixedly connected to the upper surface of the workbench 1. A conveying block 408 is threadedly connected to the outer surface of the electric lead screw 407.

[0054] The detection system 7 includes:

[0055] The image acquisition module is used to acquire images of the LED light panel;

[0056] The light source module is used to provide illumination for the image acquisition module;

[0057] The processing control module includes:

[0058] The uncertainty analysis unit is used to receive the first image acquired by the image acquisition module and calculate the confidence score and pixel-level uncertainty spectrum of the area to be inspected in the LED light board.

[0059] The active control decision unit is used to select a second set of illumination parameters from a predefined strategy library based on the characteristics of the uncertainty spectrum when the confidence score is lower than a preset threshold or the uncertainty spectrum indicates the existence of an uncertainty region, and instruct the supplementary light 606 to execute it.

[0060] The multimodal fusion decision unit is used to fuse information from multiple images acquired under different illumination parameters to obtain the final detection result of the LED light panel.

[0061] The predefined strategy library of the proactive control decision-making unit includes:

[0062] The first control strategy is to activate the LED beads with a low illumination angle in the fill light 606 when the uncertain area exhibits high light saturation characteristics, and control the rotating arm 605 to rotate the fill light 606 by a certain angle.

[0063] The second control strategy is to control the supplementary light 606 to switch between two different LED beads with different illumination angles in a short period of time when the uncertain area exhibits low contrast characteristics.

[0064] The steps for the multimodal fusion decision unit to obtain the final detection result of the LED light panel are as follows:

[0065] Based on the image pairs synchronously acquired by the main camera 603 and the auxiliary camera 604, the relative height map of the area to be inspected of the LED light panel is calculated using a stereo vision algorithm;

[0066] Images acquired under different lighting parameters and relative height maps are used as multiple evidence sources. Independent decisions are made on the same area to be inspected, and multiple local decision results are obtained. The confidence scores corresponding to each local decision result are then weighted, and the final detection result is output using a weighted voting method.

[0067] The processing control module also includes a strategy library update unit, which stores the illumination parameter strategies and corresponding uncertainty characteristics successfully executed by the active control decision unit for a specific type of LED light panel into a predefined strategy library.

[0068] Working principle: In actual use, multiple raw material boxes containing LED light panels are first stacked on the telescopic frame 201 of the feeding assembly 2, with a maximum of 20 raw material boxes. Then, the control system drives the telescopic frame 201 to lower the raw material boxes, while simultaneously driving two cylinders 203 to push the linkage plate 204 to move, so that the limit rod 206 slides inside the limiter 205 and gets stuck at the edge of the previous raw material box of the bottom raw material box. Then, the electric screw 207 rotates to drive the conveyor block 208 to push out the bottom raw material box. Afterwards, a gripper 502 on the gantry frame 501 of the transfer assembly 5 grabs the LED light panel and places it on the placement plate 303 of the dividing assembly 3. Subsequently, motor 302 of the dividing assembly 3 drives the placement plate 303 to tilt between the support columns 301 from horizontal to inclined. The LED light panel then slides down under gravity. At this time, cylinder 306 pushes the auxiliary claw 307 below the inclined opening of the placement plate 303. The auxiliary claw 307 then opens and clamps the edge of the LED light panel. Simultaneously, cylinder 304 pushes the dividing blade 305 to slide downwards inside the placement plate 303 to divide the LED light panel. After division, the placement plate 303 returns to horizontal. Cylinder 306 and auxiliary claw 307 lift the divided light panel and move it to a position on the same axis as the dividing strip 311. Then, the electric slide rail 504 outside the upright plate 503 is activated, causing the rotating block 505 to... The LED light panel on the auxiliary claw 307 is pushed onto the first dividing strip 311. The midpoint of the LED light panel on the first dividing strip 311 is located at the junction of the first dividing strip 311 and the second dividing strip 312. Then, the two cylinders 601 of the test assembly 6 push the detection electrode 602 to contact half of the LED light panel for electrical testing. During the test, the rotating arm 605 adjusts the illumination angle of the supplementary light 606, and the main camera 603 and auxiliary camera 604 acquire images. The collected data is transmitted to the processor 607, which provides the test results. During the test, when one LED light panel is fixed for testing, the second motor 309 drives the impact arc 310 to rotate around the column 308 and impact... The LED light panel is split again by striking the second dividing strip 312. After the first half of the LED light panel is tested, the rotating block 505 moves again to push the other half of the LED light panel on the second dividing strip 312 to the first dividing strip 311 for testing. After the test is completed, the transfer component 5 transfers the light panel to the storage box in the recycling component 4. After the storage box is full, the cylinder 5 403 pushes the linkage plate 2 404 to move. The limit rod 2 406 slides inside the limiter 2 405 and gets stuck at the edge of the storage box. Then the electric screw 2 407 rotates to drive the conveyor block 2 408 to transport the storage box to the storage area for the operator to take away. This completes a full division test process.

[0069] The detection system 7 described below can be referred to in correspondence with the automatic board separation tester described above.

[0070] In this embodiment, the hardware of the detection system 7 includes an image acquisition module, a light source module, and a processing and control module.

[0071] The image acquisition module consists of a main camera 603 and an auxiliary camera 604. The main camera 603 is a high-resolution color area array industrial camera used to capture the two-dimensional texture and color information of the surface of the LED light panel under test and generate a color image. The auxiliary camera 604 is a high frame rate monochrome industrial camera. The main camera 603 and the auxiliary camera 604 are fixedly mounted, and their optical axes form a fixed angle. The included angle The typical value is 10°. The two cameras are stereo-calibrated to ensure that their imaging data are in the same world coordinate system, thus providing a basis for subsequent 3D information calculations.

[0072] The light source module, serving as the supplementary light 606 in this embodiment, is driven by a rotating arm 605. The supplementary light 606 has a ring-shaped structure, consisting of three concentric rings of LED beads arranged in an inner, middle, and outer ring. The illumination angle of each ring of LED beads... They are preset to different values, for example, the inner ring. Central Outer ring The entire light source assembly is driven by a stepper motor within the rotating arm 605, allowing it to rotate around its central axis. The rotation angle is denoted as... Its range is The brightness of each ring of LED beads Both can be controlled via independent pulse width modulation (PWM) signals. Therefore, this light source provides a range of illumination angles. Rotation angle and brightness The multi-dimensional illumination parameter space constitutes .

[0073] The processor 607 is fixed to the top of the gantry 501 and integrates the detection system 7. The processor 607 establishes data communication and control connections with the main camera 603, auxiliary camera 604, rotating arm 605, and supplementary lighting 606. It is responsible for receiving image data acquired by the cameras and, according to the instructions of the internally operating detection system 7, controlling the rotating arm 605 and supplementary lighting 606 with specific illumination parameters. Provide lighting.

[0074] The detection system 7 logically includes an image acquisition module, a light source module, and a processing control module. The image acquisition module's functions are implemented by the main camera 603 and the auxiliary camera 604. The light source module's functions are implemented by the supplementary lighting 606 and the rotating arm 605. All algorithmic logic of the processing control module is executed in the processor 607.

[0075] The detection system 7, running in processor 607, operates as a closed-loop control process, with the following specific steps:

[0076] With a set of preset standard lighting parameters The control light source module provides illumination, and the image acquisition module is instructed to simultaneously acquire an initial color image. and monochrome images The processing control module processes color images. Preprocessing is performed, including image noise reduction and region of interest (ROI) localization, to determine the region of each element to be inspected. .

[0077] For each area to be inspected The uncertainty analysis unit inside the processing control module extracts its appearance feature vector. This feature vector can consist of one or more feature descriptors, such as color histograms, Hu moment invariants, and local binary pattern (LBP) texture features. The uncertainty analysis unit will extract the feature vector. Compared with standard template feature vectors pre-stored in the database Perform a comparison and calculate a confidence score. The confidence score The calculation formula is:

[0078] ;

[0079] in, This represents the L2 norm, or Euclidean distance, used to quantify the degree of difference between the current feature and the standard feature. It is a preset maximum difference threshold used to normalize the degree of difference. The closer the value is to 1, the higher the degree of matching between the current region and the standard template.

[0080] Simultaneously, the uncertainty analysis unit generates a value related to the image. Pixel-level uncertainty map at the same resolution A higher value in this graph indicates greater uncertainty in the detection result for the corresponding pixel location. It is generated by fusing the maximum values ​​of multiple uncertain factors, and its calculation formula is as follows:

[0081] ;

[0082] in, Gradient magnitude is calculated based on pixel neighborhood. Low gradient magnitude corresponds to high uncertainty value, which is used to characterize blurred or low contrast areas of the image. Highlight areas are identified by detecting whether the pixel grayscale value has reached saturation (e.g., for an 8-bit image, the grayscale value is 255). Shadow areas are identified by detecting whether the pixel grayscale value is close to zero.

[0083] Obtain the confidence scores for all regions to be inspected. Uncertainty map of the whole image Then, the system enters the active optical control decision-making step. The active control decision unit inside the processing and control module will process each... Compared with the preset confidence threshold Compare and calculate global average With uncertainty threshold Compare them. If either exists... ,or If so, the system will initiate active control.

[0084] After the control measures are initiated, the proactive control decision-making unit first analyzes the uncertainty profile. Characteristics of regions with higher median values. For example, if the high uncertainty region is mainly composed of... If the contribution is specular reflection, the diagnosis is specular reflection. If it is mainly caused by... If the contribution is insufficient, it is diagnosed as low contrast or feature blur. Based on the diagnosis, the decision unit queries a predefined mapping table of illumination strategies and uncertainty types and selects a set of optimal illumination parameters.

[0085] For example, for the problem of high light saturation caused by specular reflection, the selected strategy (first control strategy) is: to activate the fill light 606 with a low illumination angle (e.g., The inner ring of LED beads, and control the rotating arm 605 to rotate the fill light 606 by a specific angle (e.g., This is done by changing the direction of light incidence to move the highlight reflection out of the inspection area. For low contrast or feature blurring issues, the selected strategy (second control strategy) is to control the supplementary light 606 to switch between two different illumination angles sequentially within a short period of time (e.g., first using...). Reuse Simultaneously, two new sets of images were acquired, and the surface texture was enhanced by utilizing the shadow differences generated by different lighting angles.

[0086] Perform one or more active optical modulations and acquire new image sequences. Then, the system enters the multimodal information fusion decision step. The multimodal fusion decision unit within the processing control module first uses images synchronously acquired by the main camera 603 and the auxiliary camera 604 under the same illumination to... Disparity maps are calculated using a stereo matching algorithm. Then, the three-dimensional topography of the area to be inspected is reconstructed to obtain a relative height map. The calculation formula is as follows:

[0087] ;

[0088] in, The effective focal length of the camera, The baseline distance between the two cameras is the baseline distance; both parameters were obtained through prior stereo calibration. Indicates pixel coordinates The disparity value at that location is calculated by a stereo matching algorithm.

[0089] Subsequently, the multimodal fusion decision unit will process the initially acquired images. Image sequences acquired after active adjustment and the calculated relative height map As multiple independent sources of evidence. For each area to be inspected. Each source of evidence will produce an independent partial judgment. and the corresponding confidence level The final fusion ruling The weighted voting method was used to arrive at the following decision:

[0090]

[0091] in, It is an indicator function that takes the value 1 when the condition is true and 0 otherwise; weight Confidence level of this source of evidence Positive correlation; Indicates the first The final fusion judgment result of each area to be inspected is 1, which represents qualified (OK) and 0 represents unqualified (NG). This represents the total number of rounds of active data collection. For the first The second collection was performed on the first... The local judgment result generated for each area to be inspected is 1, which represents OK and 0, which represents NG.

[0092] For example, it can be set to This method ensures that local decisions with higher confidence levels carry greater weight in the final decision.

[0093] In addition, the processing control module also includes a strategy library update unit. When an active adjustment process successfully resolves an uncertainty and obtains a high-confidence final decision, this unit updates the features that caused the uncertainty (such as the area and location of the highlight region) with the lighting parameter strategies that successfully resolved the problem. Establish associations and store these association rules in a predefined strategy library for subsequent detection and optimization of specific LED light panels.

[0094] The policy library update unit within the control module is responsible for optimizing and extending the performance of the predefined policy library. This unit starts when specific conditions are met, and the specific process is as follows:

[0095] When the multimodal fusion decision unit performs final detection on paired LED light panels and outputs the decision result... Subsequently, the policy library update unit evaluates the effectiveness of this proactive adjustment process. The evaluation criteria are: if this proactive adjustment process successfully resolves the uncertainties identified in the initial detection phase, and the final decision result... The corresponding weighted total score (e.g., or Exceeding a preset final confidence threshold If so, then the regulation is considered successful.

[0096] Once the adjustment is deemed successful, the strategy library update unit extracts the original features that caused the uncertainty and the illumination parameter strategy that successfully resolved the uncertainty. The original uncertainty features include, but are not limited to, the uncertainty map. The types of medium-to-high value areas (e.g., high light saturation, low gradient blur, shadow areas), spatial distribution (e.g., centrally concentrated, edge-distributed), area ratio, and corresponding LED panel model information are required. A successful illumination parameter strategy is the one selected by the active control decision unit in this process. Parameter combinations.

[0097] The policy library update unit uses the extracted uncertainty features as keys and the corresponding successful lighting parameter policies. The corresponding values ​​are stored or updated in a predefined policy library within the active control decision-making unit. This policy library can be implemented as a hash table or a knowledge base of basic rules. Through this mechanism, the system can accumulate and learn the optimal combination of illumination parameters for specific uncertain scenarios. Thus, in future detection tasks, when encountering similar uncertain features, it can more directly and efficiently select the corresponding illumination parameter strategy, improving the adaptability of the detection.

Claims

1. An automatic board separation testing machine, comprising a workbench (1), characterized in that, A feeding component (2) is provided on one side of the upper surface of the workbench (1) for conveying undivided LED light boards. A dividing component (3) is provided in the middle of the upper surface of the workbench (1) for dividing large undivided LED light boards into smaller pieces. A recycling component (4) is provided on the other side of the upper surface of the workbench (1) for collecting the divided and tested LED light boards. A transfer component (5) is provided on the upper surface of the workbench (1) for transporting LED light boards. The transfer component (5) is located at the output end of the feeding component (2) and the dividing component (3). A testing component (6) is provided between the dividing component (3) and the recycling component (4) for detecting whether the LED light boards can be used normally. The testing component (6) integrates a detection system (7). The detection system (7) includes: The image acquisition module is used to acquire images of the LED light panel; A light source module is used to provide illumination for the image acquisition module; The processing control module includes: An uncertainty analysis unit is used to receive the first image acquired by the image acquisition module and calculate the confidence score and pixel-level uncertainty spectrum of the area to be inspected of the LED light board. An active control decision unit is used to select a second set of illumination parameters from a predefined strategy library based on the characteristics of the uncertainty spectrum when the confidence score is lower than a preset threshold or the uncertainty spectrum indicates the existence of an uncertainty region, and instruct the component to execute the selection. Among them, the pixel-level uncertainty map The calculation formula is: ； in, Gradient magnitude is calculated based on pixel neighborhood. Low gradient magnitude corresponds to high uncertainty value, which is used to characterize blurred or low contrast areas of the image. Highlight areas are identified by detecting whether the pixel grayscale value has reached saturation. Shadow areas are identified by detecting whether the pixel grayscale value is close to zero; The multimodal fusion decision unit is used to fuse information from multiple images acquired under different illumination parameters to obtain the final detection result of the LED light panel. The dividing component (3) includes two support columns (301), which are fixedly connected to the upper surface of the workbench (1). A motor (302) is fixedly connected to the outer surface of one of the support columns (301). A placement plate (303) is fixedly connected to the output end of the motor (302). The placement plate (303) is rotatably connected between the two support columns (301). A cylinder (304) is fixedly connected to the top of the placement plate (303). A dividing blade (305) is fixedly connected to the output end of the cylinder (304). The dividing blade (305) is slidably connected inside the placement plate (303).

2. The automatic board separation testing machine according to claim 1, characterized in that, The feeding assembly (2) includes a telescopic frame (201) and a support frame (202). The telescopic frame (201) and the support frame (202) are fixedly connected to one side of the upper surface of the workbench (1). Two cylinders (203) are fixedly connected to one side of the upper surface of the workbench (1). A linkage plate (204) is fixedly connected to the output end of the cylinder (203). On both sides of the cylinder (203), multiple limiters (205) are fixedly connected to one side of the upper surface of the workbench (1). A limit rod (206) is slidably connected inside the limiter (205). One end of the limit rod (206) is fixedly connected to the outer surface of the linkage plate (204). Below the support frame (202), an electric screw (207) is fixedly connected to the upper surface of the workbench (1). A conveying block (208) is threadedly connected to the outer surface of the electric screw (207).

3. The automatic board separation testing machine according to claim 1, characterized in that, The transfer assembly (5) includes a gantry (501), which is fixedly connected to the upper surface of the workbench (1). Two grippers (502) are slidably connected to the outer surface of the gantry (501). A vertical plate (503) is fixedly connected to the upper surface of the workbench (1) below the gantry (501). An electric slide rail (504) is fixedly connected to one side of the outer surface of the vertical plate (503). A rotating block (505) is slidably connected to the outer surface of the electric slide rail (504).

4. An automatic PCB separation and testing machine according to claim 3, characterized in that, The dividing assembly (3) also includes a cylinder three (306), which is fixedly connected to the upper surface of the workbench (1). The cylinder three (306) is located below the placement plate (303). An auxiliary claw (307) is fixedly connected to the output end of the cylinder three (306). A column (308) is fixedly connected to the upper surface of the workbench (1). A motor two (309) is fixedly connected to the side surface of the column (308). An impact arc (310) is fixedly connected to the output end of the motor two (309). The impact arc (310) is rotatably connected to the outside of the column (308). A dividing strip one (311) is fixedly connected to the top of the column (308). A dividing strip two (312) is rotatably connected to one end of the dividing strip one (311).

5. An automatic PCB separation and testing machine according to claim 4, characterized in that, The test assembly (6) includes two cylinders (601), which are fixedly connected to the upper surface of the workbench (1). The two cylinders (601) are located on both sides of the dividing strip (311). The output end of the cylinders (601) is fixedly connected to a detection electrode (602). The upper surface of the workbench (1) is fixedly connected to a main camera (603) and an auxiliary camera (604). The main camera (603) and the auxiliary camera (604) are located above the dividing strip (311). The upper surface of the workbench (1) is fixedly connected to a rotating arm (605). The output end of the rotating arm (605) is fixedly connected to a supplementary light (606). The test assembly (6) also includes a processor (607), which is fixedly connected to the top of the gantry (501). The detection system (7) is integrated inside the processor (607).

6. An automatic board separation testing machine according to claim 2, characterized in that, The recycling component (4) includes a telescopic frame two (401) and a support frame two (402). The telescopic frame two (401) and the support frame two (402) are fixedly connected to the other side of the upper surface of the workbench (1). Two cylinder five (403) are fixedly connected to the other side of the upper surface of the workbench (1). The output end of the cylinder five (403) is fixedly connected to a linkage plate two (404). On both sides of the cylinder five (403), a plurality of limiters two (405) are fixedly connected to one side of the upper surface of the workbench (1). The limiter two (405) is slidably connected to a limit rod two (406) inside. One end of the limit rod two (406) is fixedly connected to the outer surface of the linkage plate two (404). Below the support frame two (402), an electric screw two (407) is fixedly connected to the upper surface of the workbench (1). The outer surface of the electric screw two (407) is threadedly connected to a conveying block two (408).

7. An automatic PCB separation and testing machine according to claim 5, characterized in that, The predefined strategy library of the active control decision-making unit includes: The first control strategy is to activate the LED beads with a low illumination angle in the fill light (606) when the uncertainty area exhibits high light saturation characteristics, and control the rotating arm (605) to rotate the fill light (606) to change the incident direction of the light and move the high light reflection out of the inspection area. The second control strategy is to control the fill light (606) to switch between two different LED beads with different illumination angles in a short period of time when the uncertain area exhibits low contrast characteristics.

8. An automatic PCB separation and testing machine according to claim 5, characterized in that, The steps for the multimodal fusion decision unit to obtain the final detection result of the LED light panel are as follows: Based on the image pairs synchronously acquired by the main camera (603) and the auxiliary camera (604), the relative height map of the area to be inspected of the LED light panel is calculated by a stereo vision algorithm; Images acquired under different lighting parameters and the relative height map are used as multiple evidence sources. Independent decisions are made on the same area to be inspected, and multiple local decision results are obtained. The confidence scores corresponding to each local decision result are then weighted, and the final detection result is output using a weighted voting method.

9. An automatic PCB separation and testing machine according to claim 1, characterized in that, The processing control module also includes a strategy library update unit, which stores the illumination parameter strategies and corresponding uncertainty characteristics successfully executed by the active control decision unit for the LED light panel into the predefined strategy library.

Images