A chip screening and anti-detection system based on machine vision
Through a chip screening and detection anti-detection system based on machine vision, the problem of direction and defect identification errors in chip testing is solved. Through image processing and optimization fusion technology, the accuracy and efficiency of the test are improved to prevent chip damage.
Patent Information
- Application Number
- CN202410067963.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-01-17
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-01-17
AI Technical Summary
During the chip screening and testing process, operators are prone to incorrect device placement direction or error identification of appearance defects due to repetitive work, resulting in the device being burned out.
Design a chip screening and detection system based on machine vision, including image acquisition module, image processing module, image recognition module and target chip testing module. Through image acquisition, processing and recognition, image angle deviation is adjusted, and through optimized fusion method Improve contrast, judge the chip direction and defects, and control whether the chip is powered on.
It effectively reduces human errors caused by angle deviation and contrast problems, improves the accuracy and efficiency of chip testing, and prevents device damage.
Smart Images

Figure CN117630045B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of chip detection, and in particular relates to a chip screening and anti-detection system based on machine vision. Background Art
[0002] With the rapid development of the semiconductor industry, chips are widely used and the chip production speed is accelerating. Every day, a large number of chips need to be screened and tested through test sockets. In the chip screening and testing process, the operator identifies the device direction and appearance defects, and places it in the test socket for testing. Long-term repetitive work is prone to the phenomenon of incorrect device placement direction or incorrect appearance defect identification. If the operator does not notice and directly powers on, the device will be burned out. Summary of the invention
[0003] In view of the above-mentioned defects of the prior art, the present invention proposes a chip screening and anti-detection system based on machine vision. The technical solution designed by the present invention includes:
[0004] Image acquisition module, image processing module, image recognition module and target chip testing module;
[0005] The image acquisition module is used to obtain an image of the target chip placement position and transmit it to the image processing module;
[0006] The image processing module is used to perform a fusion operation on the image, output the fused image and transmit it to the image recognition module;
[0007] The image recognition module is used to recognize the fused image, obtain the recognition result of the target chip position signal and appearance information, and send a command to the target chip test module according to the recognition result;
[0008] The target chip test module is used to control whether to perform power-on test on the target chip according to the command.
[0009] Preferably, the image acquisition module is a BFLY-PGE-16S2M-CS black and white pixel camera with a resolution of 1440×1080 and a frame rate of 78FPS.
[0010] Specifically, if the recognition result shows that the device information is correct and the orientation is correct, a power-on test is performed; otherwise, the device is not powered on and a warning is issued to remind the operator to check the device information and orientation.
[0011] Preferably, before the image fusion operation is performed, the image is also subjected to angle deviation adjustment, and the angle deviation adjustment includes:
[0012] Construct an XY coordinate axis on the real acquisition range of the target chip test module, take the lower left corner of the real acquisition range as the coordinate origin, rightward as the positive direction of the X axis, and upward as the positive direction of the Y axis, to obtain the coordinate information of the real acquisition range;
[0013] An XY coordinate axis is constructed on the acquired image information of the target chip placement position, with the center of the image as the coordinate origin, the right as the positive direction of the X axis, and the upward as the positive direction of the Y axis, to obtain image coordinate information;
[0014] The image is rotated clockwise and counterclockwise according to the coordinate information of the real acquisition range and the image coordinate information, an image rotation matrix is constructed, and an angle deviation adjusted image is obtained through the image rotation matrix.
[0015] Preferably, the formula for rotating the image clockwise is as follows:
[0016]
[0017] Where X1 is the X coordinate of the image after clockwise rotation, and Y1 is the Y coordinate of the image after clockwise rotation. is the X-axis coordinate of the image before rotation, is the Y-axis coordinate of the image before rotation, It is the angle difference between the image coordinate information before rotation and the actual acquisition range coordinate information.
[0018] Preferably, the formula for rotating the image counterclockwise is as follows:
[0019]
[0020] Where X2 is the X coordinate of the image after counterclockwise rotation, and Y2 is the Y coordinate of the image after counterclockwise rotation. is the X-axis coordinate of the image before rotation, is the Y-axis coordinate of the image before rotation, It is the angle difference between the image coordinate information before rotation and the actual acquisition range coordinate information.
[0021] Preferably, the formula for constructing the image rotation matrix is as follows:
[0022]
[0023] Where map is the rotation matrix.
[0024] Preferably, the image fusion operation comprises the following steps:
[0025] S10: Decompose the image to obtain a corresponding high-frequency image and a low-frequency image;
[0026] S20: converting the low-frequency image corresponding to the image to be fused into a low-frequency vector matrix based on sparse representation, and fusing them to output a fused low-frequency matrix;
[0027] S30: fusing the high-frequency image and the low-frequency image corresponding to the image to be fused based on the same frequency band using the absolute value maximum strategy, and outputting a fused full-frequency matrix;
[0028] S40: Fusing the fused low-frequency matrix and the fused full-frequency matrix to output a final fused image matrix, and reconstructing the image to convert it into a final fused image.
[0029] Preferably, the step S20 of converting the low-frequency image corresponding to the image to be fused into a low-frequency vector matrix includes:
[0030] Construct a sparse representation function based on the maximum The fusion rule continuously updates the sparse coefficient vector and finally determines the The sparse coefficient vector with the largest norm, obtains the fused sparse vector through the largest sparse coefficient vector, and converts the fused sparse vector into a low-pass vector matrix;
[0031] The formula of the sparse representation function is as follows:
[0032]
[0033] In the formula, is the low-frequency image corresponding to the image to be fused, is a sparse coefficient vector, Dictionary information.
[0034] Preferably, the formula for obtaining the fused sparse vector through the largest sparse coefficient vector is as follows:
[0035]
[0036] In the formula, G is the fused sparse vector, is the maximum sparse coefficient vector, and D is the dictionary information.
[0037] Preferably, the formula for fusing the fused low-frequency matrix and the fused full-frequency matrix in step S40 is as follows:
[0038]
[0039] In the formula, is the final fused image matrix, is the low-frequency image matrix, is the full-frequency image matrix.
[0040] Beneficial effects:
[0041] 1. The present invention adjusts the angle deviation of the image through machine vision image processing, so that the width and height of the collected image are completely parallel to the width and height of the actual range that the target chip test module needs to detect, thereby reducing the image field of view limitation problem;
[0042] 2. The present invention solves the problem of low contrast in traditional chip image fusion results through an optimized fusion method, makes the judgment of direction and defect information during chip testing more efficient, and reduces human errors caused by repetitive fatigue.
[0043] 3. The present invention uses machine vision to identify and judge the direction of device placement and appearance defect information, and disconnects the power circuit of the target chip when the direction is wrong or there is a defect. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 It is a structural schematic diagram of a preferred embodiment of the present invention;
[0045] Figure 2 It is a schematic diagram of the fusion operation flow of a preferred embodiment of the present invention. DETAILED DESCRIPTION
[0046] The embodiments of the present invention are described in detail below. The following embodiments are implemented on the premise of the technical solution of the present invention, and detailed implementation methods and specific operation processes are given, but the protection scope of the present invention is not limited to the following embodiments.
[0047] The present invention proposes a chip screening and anti-detection system based on machine vision, such as Figure 1 As shown, the technical solution designed by the present invention includes:
[0048] Image acquisition module, image processing module, image recognition module and target chip testing module;
[0049] The image acquisition module is used to obtain an image of the target chip placement position and transmit it to the image processing module;
[0050] The image processing module is used to perform a fusion operation on the image, output the fused image and transmit it to the image recognition module;
[0051] The image recognition module is used to recognize the fused image, obtain the recognition result of the target chip position signal and appearance information, and send a command to the target chip test module according to the recognition result;
[0052] The target chip test module is used to control whether to perform power-on test on the target chip according to the command.
[0053] Preferably, the image acquisition module is a BFLY-PGE-16S2M-CS black and white pixel camera with a resolution of 1440×1080 and a frame rate of 78FPS.
[0054] Preferably, before the fusion operation is performed on the images, the angle deviation adjustment is also performed on the images, and the angle deviation adjustment includes:
[0055] Construct an XY coordinate axis on the actual acquisition range of the target chip test module, take the lower left corner of the actual acquisition range as the coordinate origin, rightward as the positive direction of the X axis, and upward as the positive direction of the Y axis, to obtain the coordinate information of the actual acquisition range;
[0056] An XY coordinate axis is constructed on the acquired image information of the target chip placement position, with the center of the image as the coordinate origin, the right as the positive direction of the X axis, and the upward as the positive direction of the Y axis, to obtain image coordinate information;
[0057] The image is rotated clockwise and counterclockwise according to the coordinate information of the actual acquisition range and the image coordinate information, and an image rotation matrix is constructed. The angle deviation adjustment image is obtained through the image rotation matrix.
[0058] Preferably, the formula for rotating the image clockwise is as follows:
[0059]
[0060] Where X1 is the X coordinate of the image after clockwise rotation, and Y1 is the Y coordinate of the image after clockwise rotation. is the X-axis coordinate of the image before rotation, is the Y-axis coordinate of the image before rotation, It is the angle difference between the image coordinate information before rotation and the actual acquisition range coordinate information.
[0061] Preferably, the formula for rotating the image counterclockwise is as follows:
[0062]
[0063] Where X2 is the X coordinate of the image after counterclockwise rotation, and Y2 is the Y coordinate of the image after counterclockwise rotation. is the X-axis coordinate of the image before rotation, is the Y-axis coordinate of the image before rotation, It is the angle difference between the image coordinate information before rotation and the actual acquisition range coordinate information.
[0064] Preferably, the formula for constructing the image rotation matrix is as follows:
[0065]
[0066] Where map is the rotation matrix.
[0067] Specifically, since the acquisition module will have a certain angle deviation after installation, the width and height of the captured image are not completely parallel to the width and height of the actual range of the target chip detection module. Therefore, the image obtained by the acquisition module is rotated and adjusted, and the rotation center is set to the center point of the actual image, that is, , due to the use of a camera with a resolution of 1440×1080, , set a two-row and three-column rotation matrix to correct the image information obtained by the camera. Multiply the image information obtained by the camera by the rotation matrix to obtain the rotated image information.
[0068] Specifically, the traditional method is to decompose the source image separately to obtain low-frequency and high-frequency information, and then fuse the low-frequency and high-frequency information according to the corresponding fusion rule, and finally obtain the final fused image by using the fused high- and low-frequency information through the corresponding inverse transformation, and basically adopt the average fusion rule. Therefore, the contrast of the target will be reduced, because most of the energy of the picture is contained in the low frequency, and the average fusion rule will lose the energy in the source image. Therefore, the present invention adopts an optimized fusion method to optimize the fusion of low-frequency and high-frequency images, and after the angle deviation is adjusted and before the fusion operation, the image is subjected to enhanced contrast processing.
[0069] First, the source image is normalized (removing the dimension to facilitate subsequent calculations) and filtered using a guided filter to obtain the filtered image, which is then logarithmically transformed. The logarithmically transformed image is the base layer, and the base layer is obtained (the base layer is the overall brightness and color information in the image), and the detail layer is obtained based on the base layer (the base layer is obtained by subtracting the base layer from the source image, and the detail layer is the subtle structure and texture details in the image). The target contrast is defined to reduce the contrast of the base layer, and then the scale factor and recovery factor are set (the contrast of the base layer is reduced by the scale factor. The scale factor is usually a value less than 1, which controls the degree of contrast reduction. The recovery factor is a positive value used to restore the contrast on the base layer). Dynamic range compression is performed on the base layer (dynamic range compression is achieved by multiplying the pixel value in the base layer by the scale factor, which will make the originally large pixel value smaller, thereby reducing the brightness difference and reducing the contrast) and contrast recovery (the base layer compressed by the dynamic range is divided by the recovery factor to restore the original contrast level). Since the dynamic range is compressed, the overall contrast increases, and after the contrast is restored, the overall contrast is improved, thereby enhancing the contrast of the original image, making the image details clearer and the brightness difference more obvious.
[0070] Normalization processing includes normalizing the source output image to Range, the formula is:
[0071]
[0072] In the formula,
[0073] The image information after normalization processing,
[0074] is the original image information;
[0075] The filtered image formula is:
[0076]
[0077] In the formula, is the image information after filtering, GF is the guided filtering function, r is the size of GF, is the edge preservation degree of GF. The value range of r in the present invention is set in An integer between , a smaller r value can better preserve image details but may increase the effect of noise, and a larger r value will filter the image more smoothly. Controls the degree to which the guide filter maintains the edge. The present invention sets The smaller the value between Higher values will more strongly preserve edges, but may result in some distortion or poor filtering results.
[0078] The formula for the base layer is:
[0079]
[0080] Where, LFC is the base function;
[0081] The formula for the detail layer is:
[0082]
[0083] Where HFC is the detail layer function
[0084] The guided filter (GF) is a good edge protection filtering method. It uses the local linear relationship between the guided image and the filtered output image to perform the least squares method on the calculated output image and the filtered image. The output image and the guided image are locally a linear model.
[0085] Dynamic range compression and contrast restoration on the base layer are expressed by the following formula:
[0086]
[0087] In the formula, is the basic state function, is the scale factor, is the recovery factor;
[0088] and The calculation methods are as follows:
[0089]
[0090]
[0091] In the formula, For the defined contrast, for The maximum value of for The minimum value of T is defined according to the requirements. The contrast target T of the present invention is selected in the range of A smaller T value will reduce the contrast of the image, and a larger T value will increase the contrast of the image.
[0092] The enhanced image information is expressed as follows:
[0093]
[0094] The above method enhances the chip feature information and improves the accuracy of image information, and then fuses the image information after enhancing the contrast.
[0095] The same frequency band of low frequency and high frequency is merged, specifically and , and , and , and Perform absolute value maximum strategy fusion to obtain the full-pass fusion matrix The absolute value maximum strategy is to fuse the two images using the absolute value maximum strategy, compare the absolute values of the image information on the two images, select the largest one as the fusion result, and apply the sliding window technology, using a 3×3 sliding window to slide on the matrix to obtain the feature information of the local area.
[0096] The preferred specific embodiments of the present invention are described in detail above. It should be understood that ordinary technicians in the field can make many modifications and changes based on the concept of the present invention without creative work. Therefore, all technical solutions that can be obtained by technicians in the technical field based on the concept of the present invention through logical analysis, reasoning or limited experiments on the basis of the prior art should be within the scope of protection determined by the claims.
Claims
1. A chip screening and anti-detection system based on machine vision, characterized in that: include: Image acquisition module, image processing module, image recognition module and target chip testing module; The image acquisition module is used to obtain an image of the target chip placement position and transmit it to the image processing module; The image processing module is used to perform a fusion operation on the image, output the fused image and transmit it to the image recognition module; The image recognition module is used to recognize the fused image, obtain the recognition result of the target chip position signal and appearance information, and send a command to the target chip test module according to the recognition result; The target chip test module is used to control whether to conduct a power-on test on the target chip according to the command; Before the image fusion operation is performed, the image is further subjected to angle deviation adjustment, and the angle deviation adjustment includes: Construct an XY coordinate axis on the real acquisition range of the target chip test module, take the lower left corner of the real acquisition range as the coordinate origin, rightward as the positive direction of the X axis, and upward as the positive direction of the Y axis, to obtain the coordinate information of the real acquisition range; An XY coordinate axis is constructed on the acquired image information of the target chip placement position, with the center of the image as the coordinate origin, the right as the positive direction of the X axis, and the upward as the positive direction of the Y axis, to obtain image coordinate information; According to the coordinate information of the real acquisition range and the image coordinate information, the image is rotated clockwise and counterclockwise to construct an image rotation matrix, and an angle deviation adjustment image is obtained through the image rotation matrix; The image fusion operation comprises the following steps: S10: Decompose the image to obtain a corresponding high-frequency image and a low-frequency image; S20: converting the low-frequency image corresponding to the image to be fused into a low-frequency vector matrix based on sparse representation, and fusing them to output a fused low-frequency matrix; S30: fusing the high-frequency image and the low-frequency image corresponding to the image to be fused based on the same frequency band using the absolute value maximum strategy, and outputting a fused full-frequency matrix; S40: fusing the fused low-frequency matrix and the fused full-frequency matrix to output a final fused image matrix, and reconstructing the image to convert it into a final fused image; The step S20 of converting the low-frequency image corresponding to the image to be fused into a low-frequency vector matrix includes: Construct a sparse representation function, based on the maximum fusion rule, continuously update the sparse coefficient vector and finally determine the sparse coefficient vector with the largest norm under the condition of satisfying the low-frequency image constraint. Obtain the fused sparse vector through the largest sparse coefficient vector, and convert the fused sparse vector into a low-pass vector matrix. The formula of the sparse representation function is as follows: In the formula, is the low-frequency image corresponding to the image to be fused, is a sparse coefficient vector, For dictionary information; The formula for obtaining the fused sparse vector through the largest sparse coefficient vector is as follows: In the formula, G is the fused sparse vector, is the maximum sparse coefficient vector, and D is the dictionary information; The formula for fusing the fused low-frequency matrix and the fused full-frequency matrix in step S40 is as follows: In the formula, is the final fused image matrix, is the low-frequency image matrix, is the full-frequency image matrix.
2. According to the machine vision-based chip screening and anti-detection system of claim 1, it is characterized in that: The image acquisition module is a BFLY-PGE-16S2M-CS black and white pixel camera with a resolution of 1440×1080 and a frame rate of 78FPS.
3. The chip screening and anti-detection system based on machine vision according to claim 1, characterized in that: The formula for rotating the image clockwise is as follows: Where X1 is the X coordinate of the image after clockwise rotation, Y1 is the Y coordinate of the image after clockwise rotation, is the X-axis coordinate of the image before rotation, is the Y-axis coordinate of the image before rotation, and is the angle difference between the image coordinate information before rotation and the coordinate information of the actual acquisition range.
4. The chip screening and anti-detection system based on machine vision according to claim 1, characterized in that: The formula for rotating the image counterclockwise is as follows: Where X2 is the X coordinate of the image after counterclockwise rotation, and Y2 is the Y coordinate of the image after counterclockwise rotation. is the X-axis coordinate of the image before rotation, is the Y-axis coordinate of the image before rotation, It is the angle difference between the image coordinate information before rotation and the actual acquisition range coordinate information.
5. A chip screening and anti-detection system based on machine vision according to any one of claims 3-4, characterized in that: The formula for constructing the image rotation matrix is as follows: Where map is the rotation matrix.
