A mobile visual quality inspection device for production lines

Abstract

This invention belongs to the field of visual inspection technology, specifically relating to a mobile visual quality inspection device for production lines. The invention employs a triple-guided limiting structure consisting of an upper guide rail, a lower guide rail, horizontal pulleys, vertical pulleys, and a slider. This achieves smooth, unobstructed movement of the AGV's mobile base plate, providing a stable mobile imaging platform for the visual inspection system. Simultaneously, the real-time position data of the structure is synchronously transmitted to the inspection system. The system can preset different defect detection sub-modules and dimensional measurement sub-models for different product types. Based on the information of the moved workstation, it can automatically call upon the corresponding workstation's detection model and calibration parameters, achieving adaptive inspection across multiple workstations throughout the production line. This eliminates the need to deploy multiple independent devices, significantly reducing the deployment cost of multi-workstation quality inspection and solving the problem of inflexible deployment in traditional fixed-structure systems.

Description

Technical Field

[0001] This invention belongs to the field of visual inspection technology, specifically relating to a mobile production line visual quality inspection device. Background Technology

[0002] With the development of industrial automation, assembly line visual inspection technology has become a core means of product quality control. It uses a mechanical structure to carry a visual inspection system to achieve automated detection of product appearance defects and dimensional parameters.

[0003] However, in existing technologies, mechanical structures and visual inspection systems are mostly designed independently and lack coordination, resulting in inaccurate overall inspection performance. To address this, a mobile, automated visual quality inspection device has been designed. Summary of the Invention

[0004] In view of the above-mentioned shortcomings in the prior art, the present invention provides a mobile assembly line visual quality inspection device to solve the problems in the background art.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: A mobile visual quality inspection device for production lines includes symmetrically arranged frames with conveyor belts between them. Upper and lower guide rails are connected to the upper and lower parts of each frame, respectively. Both the upper and lower guide rails are equipped with first wheel rails. An AGV (Automated Guided Vehicle) mobile base plate is located between the upper and lower guide rails. The AGV mobile base plate has horizontal pulleys that are slidably connected to the first wheel rails. A connecting column is connected to one end of the AGV mobile base plate, and a first connecting plate is connected to the connecting column. A first mounting plate is connected to the upper end of the first connecting plate, and a control plate is connected to the upper end of the first mounting plate. A data acquisition component is located at the lower end of the first mounting plate. The device also includes auxiliary fixing components connected to the AGV mobile base plate.

[0006] Preferably, the acquisition component includes an area array camera, a first electric push rod is connected to the upper end of the first mounting plate, the output shaft of the first electric push rod is connected to a second mounting plate, and the area array camera is connected to the second mounting plate.

[0007] Preferably, the second mounting plate is connected to sliding rods at both ends, the first mounting plate is provided with through holes, the sliding rods are slidably connected to the through holes, and a connecting block is connected to the upper end of the sliding rods, the width of the connecting block being greater than the width of the through holes.

[0008] Preferably, the auxiliary fixing component includes a friction block, the AGV moving base plate is connected to a second connecting plate, the second connecting plate is connected to a second electric push rod, and the output shaft of the second electric push rod is connected to the friction block.

[0009] Preferably, a second fixed plate is connected to the side of the frame near the AGV moving base plate, and a pressure sensor is connected to the second fixed plate. A friction surface layer is provided at the end of the pressure sensor away from the second fixed plate, and the friction surface layer cooperates with the friction block.

[0010] Preferably, the lower end of the AGV mobile base plate is connected to a first fixed plate, the lower end of the first fixed plate is connected to a vertical pulley, and a second wheel rail is provided on the inner side of the lower guide rail, with the vertical pulley slidably connected to the second wheel rail.

[0011] Preferably, the upper end of the AGV mobile base plate is connected to a slider, and the inner side of the upper guide rail is provided with a sliding groove, and the slider is slidably connected to the sliding groove.

[0012] Preferably, the friction surface layer has a separator ball on the side away from the pressure sensor, and the separator ball is used to divide the pressure sensor into sections.

[0013] A mobile production line visual quality inspection system includes an acquisition unit, a preprocessing unit, a processing unit, and a communication unit. The acquisition unit is used to acquire images of products on the production line; The preprocessing unit is used to perform deblurring preprocessing on the acquired images; The processing unit is used to extract features, identify defects, and measure dimensions of the preprocessed image, and output the quality inspection results in real time. The communication unit is used to transmit the quality inspection results.

[0014] Preferably, the preprocessing unit includes an exposure duration control unit and an image compensation control unit. The exposure duration control unit is used to calculate the motion displacement of the product based on the operating speed of the production line, and generate an exposure duration control signal and image compensation parameters based on the motion displacement. The exposure duration control signal is used to control the exposure time of the high-speed area scan camera to be less than the product motion time corresponding to a single pixel. The image compensation control unit is used to perform deblurring preprocessing on the acquired image based on the image compensation parameters.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention can accurately sense motion parameters such as production line speed and product position, and synchronize these parameters to the vision inspection system in real time, providing accurate input basis for the system's dynamic exposure and motion blur compensation. The system can dynamically adjust the camera's exposure time according to these parameters to ensure that the product displacement is less than a single pixel within the exposure period. At the same time, it calculates the motion blur kernel by combining accurate motion parameters, and achieves high-precision motion blur compensation through Wiener filtering algorithm, which greatly reduces the image blur problem caused by the movement of production line products, greatly improves the image clarity and detail reproduction, and provides a high-quality imaging foundation for defect identification and size measurement. 2. The auxiliary fixing component of this invention adopts a rigid locking structure that combines friction blocks and pressure sensors. This not only achieves precise rigid locking after the device is moved into position, effectively resisting vibration interference from the production line, but also provides real-time pressure data from the pressure sensor to the control system. The system can determine the locking status based on the pressure data: the system only starts the detection process when the pressure on both sides is uniform and the device is locked in place, avoiding false detections in the unlocked state. Furthermore, the system can monitor the vibration amplitude of the device in real time based on the pressure data and dynamically adjust the image noise reduction parameters of the vision system, effectively eliminating imaging noise caused by vibration and significantly improving the stability and reliability of the detection. 3. This invention employs a triple-guided limiting structure consisting of an upper guide rail, a lower guide rail, a horizontal pulley, a vertical pulley, and a slider. This achieves smooth, sway-free, and jam-free movement of the AGV's mobile base plate, providing a stable mobile imaging platform for the vision inspection system. Simultaneously, the real-time position data of the structure is synchronously transmitted to the inspection system. The system can preset different defect detection sub-modules and dimensional measurement sub-models for different product types. Based on the information of the moved workstation, it can automatically call the corresponding workstation's detection model and calibration parameters, realizing adaptive inspection across multiple workstations in the entire production line. This eliminates the need to deploy multiple independent devices, significantly reducing the deployment cost of multi-workstation quality inspection and solving the problem of inflexible deployment in traditional fixed structure systems. Attached Figure Description

[0016] Figure 1 This is a three-dimensional structural diagram of the present invention; Figure 2 This is a schematic diagram of the AGV mobile base plate structure provided by the present invention; Figure 3 The frame and guide rail structure diagram provided for this invention; Figure 4 A schematic diagram of the first mounting plate structure provided by the present invention. Figure 5 A processing system diagram provided by the present invention; The reference numerals in the accompanying drawings include: frame 1, conveyor belt 2, support column 3, upper guide rail 4, AGV moving base plate 5, first connecting plate 6, first mounting plate 7, control plate 8, first electric push rod 9, second mounting plate 10, area array camera 11, second connecting plate 12, second electric push rod 13, friction block 14, connecting column 15, horizontal pulley 16, slider 17, first fixed plate 18, vertical pulley 19, lower guide rail 20, second fixed plate 21, pressure sensor 22, slide groove 23, first wheel rail 24, separator ball 25, second wheel rail 26, sliding rod 27, connecting block 28, friction surface layer 29. Detailed Implementation

[0017] To enable those skilled in the art to better understand the present invention, the technical solution of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0018] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual images. They should not be construed as limiting the scope of this patent. To better illustrate the embodiments of the present invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions of the product. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0019] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "inner," and "outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present patent. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0020] In the description of this invention, unless otherwise explicitly specified and limited, the term "connection" or similar designation indicating a connection between components should be interpreted broadly. For example, it can refer to a fixed connection, a detachable connection, or an integral part; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can refer to the internal communication between two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0021] Example 1: like Figure 1-5The illustrated mobile visual quality inspection device for production lines includes symmetrically arranged frames 1, with conveyor belts 2 between the frames 1. Upper guide rails 4 and lower guide rails 20 are respectively connected to the upper and lower parts of the frames 1. Both the upper guide rails 4 and lower guide rails 20 are equipped with first wheel rails 24. An AGV mobile base plate 5 is located between the upper guide rails 4 and lower guide rails 20. The AGV mobile base plate 5 is equipped with horizontal pulleys 16, which are slidably connected to the first wheel rails 24. A connecting column 15 is connected to one end of the AGV mobile base plate 5, and a first connecting plate 6 is connected to the connecting column 15. A first mounting plate 7 is connected to the upper end of the first connecting plate 6, and a control plate 8 is connected to the upper end of the first mounting plate 7. A data acquisition component is located at the lower end of the first mounting plate 7. The device also includes auxiliary fixing components connected to the AGV mobile base plate 5.

[0022] Furthermore, the acquisition component includes an area array camera 11. A first electric push rod 9 is connected to the upper end of the first mounting plate 7. The output shaft of the first electric push rod 9 is connected to a second mounting plate 10. The area array camera 11 is connected to the second mounting plate 10. Sliding rods 27 are connected to both ends of the second mounting plate 10. The first mounting plate 7 has through holes. The sliding rods 27 are slidably connected to the through holes. A connecting block 28 is connected to the upper end of the sliding rod 27. The width of the connecting block 28 is greater than the width of the through hole. The auxiliary fixing component includes a friction block 14. A second connecting plate 12 is connected to the AGV moving base plate 5. A second electric push rod 13 is connected to the second connecting plate 12. The output shaft of the second electric push rod 13 is connected to the friction block 14. The frame 1 is close to the AGV moving base plate 5. A second fixed plate 21 is connected to one side of the AGV mobile base plate 5, and a pressure sensor 22 is connected to the second fixed plate 21. A friction surface layer 29 is provided at the end of the pressure sensor 22 away from the second fixed plate 21. The friction surface layer 29 cooperates with the friction block 14. A first fixed plate 18 is connected to the lower end of the AGV mobile base plate 5, and a vertical pulley 19 is connected to the lower end of the first fixed plate 18. A second wheel rail 26 is provided on the inner side of the lower guide rail 20, and the vertical pulley 19 is slidably connected to the second wheel rail 26. A slider 17 is connected to the upper end of the AGV mobile base plate 5, and a sliding groove 23 is provided on the inner side of the upper guide rail 4. The slider 17 is slidably connected to the sliding groove 23. A separator ball 25 is provided on the side of the friction surface layer 29 away from the pressure sensor 22. The separator ball 25 is used to divide the pressure sensor 22 into sections.

[0023] A mobile production line visual quality inspection system includes an acquisition unit, a preprocessing unit, a processing unit, and a communication unit. The acquisition unit is used to acquire images of products on the production line; The preprocessing unit is used to perform deblurring preprocessing on the acquired images; The processing unit is used to extract features, identify defects, and measure dimensions of the preprocessed image, and output the quality inspection results in real time. The communication unit is used to transmit the quality inspection results.

[0024] Furthermore, the preprocessing unit includes an exposure duration control unit and an image compensation control unit. The exposure duration control unit is used to calculate the motion displacement of the product based on the operating speed of the production line, and generate an exposure duration control signal and image compensation parameters based on the motion displacement. The exposure duration control signal is used to control the exposure time of the high-speed area array camera 11 to be less than the product motion time corresponding to a single pixel. The image compensation control unit is used to perform deblurring preprocessing on the acquired image based on the image compensation parameters.

[0025] Working principle: This embodiment discloses a mobile visual quality inspection device for an assembly line. The core includes two symmetrically arranged frames 1, which are arranged in parallel to form the main support structure of the assembly line. A conveyor belt 2 for transporting products to be inspected is provided between the two frames 1. The conveyor belt 2 is connected to a drive motor via transmission rollers to achieve continuous transport of the products to be inspected. An upper guide rail 4 and a lower guide rail 20 are fixedly connected to the upper inner side and lower inner side of the two frames 1, respectively. Both the upper guide rail 4 and the lower guide rail 20 are along the length direction of the assembly line (i.e., the transport direction of the conveyor belt 2). Extended, with the upper guide rail 4 and lower guide rail 20 arranged parallel and directly opposite each other, the frame 1 has first wheel rails 24 on both the upper and lower sides, and an AGV moving base plate 5 is provided between the upper guide rail 4 and the lower guide rail 20. Two horizontal rollers 16 are connected to the upper and lower right sides of the AGV moving base plate 5, and the two horizontal rollers are rotatably connected to the first wheel rails 24 on the frame 1. The wheel surface of the horizontal rollers 16 is in contact with the inner side wall of the first wheel rails 24, so as to realize the movement of the AGV moving base plate 5 along the length of the assembly line, while limiting the movement of the AGV moving base plate 5 along the width of the assembly line. To prevent lateral swaying and further improve the stability of the AGV mobile base plate 5, a first fixed plate 18 is fixedly connected to the lower end of the AGV mobile base plate 5. A vertical pulley 19 is connected to the lower end of the first fixed plate 18. A second wheel rail 26 is provided on the inner wall of the lower guide rail 20. The second wheel rail 26 runs through the length of the assembly line, and its direction is perpendicular to the first wheel rail 24. The wheel surface of the vertical pulley 19 is in rolling contact with the bottom surface of the second wheel rail 26, thus providing a smooth flow path for the AGV mobile base plate 5 along the length of the assembly line. The upper end of the AGV mobile base plate 5 is fixedly connected to an upwardly extending slider 17. The inner side wall of the upper guide rail 4 is provided with a groove 23 that runs through the length of the assembly line. The slider 17 is embedded in the groove 23 and slidably connected to the groove 23. Through the triple guide and limiting structure formed by the horizontal pulley 16 and the first wheel rail 24, the vertical pulley 19 and the second wheel rail 26, and the slider 17 and the groove 23, the AGV mobile base plate 5 is guaranteed to be stable throughout the movement without jamming or swaying. A first electric push rod 9 is symmetrically provided at the upper end of the first mounting plate 7. Its output shaft passes vertically downward through the first mounting plate 7, and the end of the output shaft is fixedly connected to a horizontally arranged second mounting plate 10. The area scan camera 11 is fixedly mounted on the lower end face of the second mounting plate 10 through a camera bracket, and the lens of the area scan camera 11 faces vertically toward the upper surface of the conveyor belt 2. L-shaped sliding rods 27 are fixedly connected to both the left and right ends of the second mounting plate 10. Two through holes are opened on the first mounting plate 7 at the vertically upward positions corresponding to the two sliding rods 27. The outer wall of the sliding rod 27 is clearance-fitted with the inner wall of the through hole to realize the free sliding of the sliding rod 27 along the through hole. A circular connecting block 28 is fixedly connected to the upper end of the sliding rod 27. The diameter of the connecting block 28 is larger than the diameter of the through hole, which can limit the downward limit position of the sliding rod 27 and prevent the area scan camera 11 from colliding with the product to be inspected. The AGV mobile base plate 5 is also connected to an auxiliary fixing component for rigid locking after the device is in place. Specifically, the auxiliary fixing component includes a friction block 14 made of wear-resistant rubber. A second connecting plate 12 is fixedly connected to both the left and right end faces of the AGV mobile base plate 5 along the width direction of the production line. A second electric push rod 13 is fixedly installed on one side of the second connecting plate 12. The output end of the second electric push rod 13 passes through the second connecting plate 12 and is fixedly connected to the friction block 14. A second fixing plate 21 is fixedly connected to the inner sidewalls of both sets of frames 1. The second fixing plate 21 extends along the length direction of the production line. A membrane-type... Pressure sensor 22 is covered with a friction surface layer 29 made of wear-resistant rubber. The pressure sensor 22 is divided into sections by a separator ball 25. When the AGV moving base plate 5 moves to the production line, the second electric push rod 13 can be activated to drive the friction block 14 to squeeze the pressure sensor 22. Then, by receiving the signal from the pressure sensor 22, it can help determine whether the position of the AGV moving base plate 5 is within a suitable range and fix the area array camera 11. The extension amount of the second electric push rods 13 on both sides can be adjusted according to the data of the pressure sensor 22 to ensure that the pressure on both sides is sufficient to keep the area array camera 11 stable. The control board 8 also integrates a visual quality inspection system, including an acquisition unit, a preprocessing unit, a processing unit and a communication unit. The acquisition unit is electrically connected to the area array camera 11 and the product arrival photoelectric sensor of the production line. When the product to be inspected arrives, the acquisition unit controls the area array camera 11 to trigger the taking of pictures, acquires the original surface image of the product to be inspected on the conveyor belt 2, and transmits the original image to the preprocessing unit. The preprocessing unit is signal-connected to the acquisition unit and is used to perform deblurring preprocessing on the acquired raw image to improve the clarity of image details. Specifically, the exposure time control unit is electrically connected to the rotary encoder and the drive circuit of the area scan camera 11 of the production line. The rotary encoder collects the running speed signal of the conveyor belt 2 in real time and converts it into the real-time linear speed of the conveyor belt 2. Unit: mm / s, and the velocity value is uploaded to the exposure time control unit in real time. The exposure time control unit adjusts the exposure time based on the real-time linear velocity of conveyor belt 2. Calculate the displacement of the product under inspection during the camera exposure time. in To preset the initial exposure time, the exposure time control unit retrieves the hardware parameters of the area scan camera 11, including the physical size of a single pixel. (Unit: mm / pixel) Camera field of view width (Unit: mm) Image horizontal resolution (Unit: pixel) Calculate the actual physical equivalent of a single pixel based on the field of view and resolution. , That is, a single pixel represents the surface of the product. The actual size of the product is determined, and then the movement time required for the product to move one pixel is calculated: ,Should This refers to the product movement time corresponding to a single pixel. Based on the above calculation results, the exposure time control unit generates an exposure time control signal to adjust the actual exposure time of the area scan camera 11. satisfy This ensures that the product's movement displacement in the image does not exceed one pixel within a single exposure cycle; Meanwhile, the exposure duration control unit adjusts according to the real-time linear velocity. Exposure time Camera lens magnification A motion fuzzy kernel model is established for the product's motion direction. The calculation process is as follows: Determine the direction angle of motion The angle between the product's movement direction and the horizontal axis of the image is calculated using the rotation signal from the rotary encoder on the production line and the preset value of the camera's mounting angle. If the product moves horizontally along the image, the angle is 0°; if it moves vertically, the angle is 90°. The calculation formula is as follows: in The velocity component of the product along the horizontal direction of the image. The velocity component of the product along the vertical direction of the image. , ,in Install preset angles on the camera; Calculate the fuzzy kernel radius The calculation formula is: The calculation result is rounded down to an integer and used as the half-width value of the fuzzy kernel; Calculate the fuzzy intensity coefficient The blur intensity is positively correlated with linear velocity and exposure time, and negatively correlated with lens magnification. The calculation formula is as follows: , The value ranges from 0.1 to 1.0. The larger the value, the stronger the blurring effect. It is used for weight adjustment in the subsequent Wiener filtering algorithm. The exposure time control unit will calculate the motion direction angle. Fuzzy kernel radius Fuzzy intensity coefficient The camera distortion coefficient and lens focal length are combined to form a complete set of image compensation parameters, which are then transmitted to the image compensation control unit. The image compensation control unit constructs a point spread function (PSF) for linear motion blur based on the motion direction angle and the blur kernel radius. The size of the PSF is... The function expression is: ,in Let the pixel coordinates satisfy If the value is outside the range, it is taken as 0; Based on the constructed PSF and fuzzy intensity coefficient The Wiener filtering algorithm is used to deblur and restore the original image. The algorithm formula is as follows: ,in This is the frequency domain representation of the deblurred image. For the frequency domain representation of PSF, for The conjugate of complex numbers, This is the frequency domain representation of the original blurred image. This is the blur intensity coefficient, which is used to quickly eliminate linear motion blur and restore image details. Subpixel-level edge optimization is performed on the deblurred image using the Canny operator combined with a parabolic fitting algorithm. The specific steps are as follows: First, the coarse edges of the product are extracted using the Canny operator to obtain the edge pixel coordinates; then, a grayscale parabolic fitting is performed on the 3×3 pixel area near the coarse edges, and the vertex coordinates of the fitted curve are calculated. These vertex coordinates are the subpixel-level edge coordinates. Finally, offset corrections are performed on key edges such as the product outline, corners, and holes to eliminate residual subpixel-level blur after hardware exposure control, improve edge contrast and positional accuracy, ensure the accuracy of subsequent defect identification and size measurement, and provide a high-quality image foundation for the subsequent processing unit.

[0026] The processing unit receives images with subpixel-level edge optimization. The unit includes a built-in defect detection submodule and a size measurement submodel. The pre-trained defect detection submodel performs quality inspection on the product. This submodel is trained based on the YOLOv8 object detection algorithm. Specifically: Training set construction: Collect at least 10,000 images of the products to be inspected under different types, angles, and lighting conditions, including eight typical appearance defects such as scratches, stains, damage, missing materials, deformation, color difference, burrs, and dents. The number of samples for each defect type should be no less than 1,000. Data augmentation processing is performed on the samples, including random rotation (0°~360°), scaling (0.8~1.2x), flipping, Gaussian noise addition, and brightness adjustment (±20%). Expand the training set to 20,000 images. The tool annotates the defect area in VOC format, clearly defining the defect category and the coordinates of the defect bounding box, and divides it into training set, test set and validation set in a 7:2:1 ratio; Model training parameters: Input image size set to 640×640 pixels, batch size set to 32, initial learning rate set to 0.001, cosine annealing learning rate decay strategy used, training epochs set to 100, and cross-entropy loss function used. The loss function stops training when the accuracy on the validation set does not improve for 10 consecutive rounds. After the preprocessed image is input into the defect detection sub-model, the model first passes through the backbone network. Multi-scale convolutional feature extraction is performed on the image to obtain feature maps at different levels (shallow feature maps are used to detect small defects, and deep feature maps are used to detect large defects); then the image is processed through a neck network ( The system fuses multi-scale feature maps to enhance feature representation capabilities. Finally, a head network performs bounding box regression and category prediction, outputting relevant parameters for all defects in the image, including: defect category (e.g., scratches, stains), defect confidence (values ​​from 0 to 1, with a confidence level ≥ 0.5 indicating a valid defect), defect pixel coordinates (the top-left and bottom-right pixel coordinates of the defect bounding box), and defect outer box pixel dimensions (width × height). Subsequently, the camera-calibrated pixel equivalent is retrieved. The formula is: Actual defect size = Defect bounding box pixel size × The pixel size is converted into the actual physical size (unit: mm) to obtain the actual size of the defect, providing data support for subsequent acceptance judgment.

[0027] The dimensional measurement sub-model, based on a sub-pixel edge detection algorithm, is used to achieve high-precision measurement of various key dimensions of the product under inspection. The specific implementation steps and calculation process are as follows: First, the deblurred image output from the preprocessing unit undergoes secondary optimization processing. A weighted average method is used to convert the color image to a grayscale image, with the following formula: (in , , These represent the grayscale values ​​of the red, green, and blue channels of the color image, respectively. Color interference is eliminated, and then a Gaussian filtering algorithm is applied, with the filter kernel size set to 5×5 and the standard deviation... =1.2, the grayscale image is smoothed to eliminate image noise. Finally, the Otsu algorithm is used to calculate the binarization threshold, converting the grayscale image into a black and white binary image, where the product area is the foreground (white, grayscale value 255) and the background area is black (grayscale value 0). Then, sub-pixel edges are extracted, and the Canny operator is used to extract the coarse edges of the product. Next, the image gradient magnitude and direction are calculated. Specifically, Sobel convolution kernels are used to perform convolution operations with the grayscale image to obtain the gradient components in the horizontal and vertical directions. Horizontal gradient kernel ; Vertical gradient kernel ; For each pixel in the image Perform convolution calculations separately: ; ; in This is the grayscale value of the current pixel. For horizontal gradient components, This represents the vertical gradient component.

[0028] Then calculate the amplitude using the following formula: The larger the calculated amplitude, the more drastic the change in grayscale value, and the more likely it is to be the edge of the product. The gradient direction is the edge normal direction, resulting in an angle range of... The continuous angle is quantized into four discrete principal directions: 0° (corresponding angle ranges: −22.5° to +22.5°, 157.5° to 180°, −180° to −157.5°), 45° (corresponding angle ranges: 22.5° to 67.5°, −157.5° to −112.5°), 90° (corresponding angle ranges: 67.5° to 112.5°, −112.5° to −67.5°), and 135° (corresponding angle ranges: 112.5° to 157.5°, −67.5° to −22.5°). Only one of these angles is retained for each pixel. A main direction is used for non-maximum suppression along the gradient direction, retaining pixels with the largest local gradient magnitude as candidate edge points; after gradient calculation and non-maximum suppression, edge pixels with gradient magnitudes greater than the high threshold (preset to 200) are retained, as are pixels with gradient magnitudes between the high and low thresholds (preset to 100) that are connected to the high threshold edge, thus obtaining the coordinates of the coarse edge pixels; sub-pixel fitting optimization: a parabolic grayscale fitting is performed on the 3×3 pixel region near the coarse edge, taking the grayscale data within the region centered on the edge pixel, and constructing a two-dimensional parabolic fitting model: in (These are the fitting coefficients). The fitting coefficients are solved using the least squares method, and the vertex coordinates of the fitted curve are calculated. These vertex coordinates are the sub-pixel level edge coordinates. The dimensional measurement sub-model retrieves pre-completion camera calibration parameters, including camera intrinsic parameters such as focal length. pixel coordinates Pixel aspect ratio Camera extrinsic parameters: rotation matrix (A 3×3 matrix describing the rotation relationship between the camera coordinate system and the world coordinate system), translation vector (3×1 vector, describing the translation relationship between the camera coordinate system and the world coordinate system), distortion coefficients: radial distortion coefficients , , Tangential distortion coefficient , , Then, combining subpixel edge coordinates, camera calibration parameters, and pixel physical equivalents... The key dimensions of the product, such as length, width, height, hole diameter, hole spacing, contour, and flatness, are calculated and fitted using coordinates to obtain the actual measured values. The specific calculation method is as follows: Length and width measurement: Extract the sub-pixel edges of two opposite sides in the length / width direction of the product, fit two parallel straight lines to each edge, and calculate the perpendicular distance between the two lines, which is the actual length / width D of the product. The calculation formula is: ,(in , (The equations of the fitted lines for the two opposite sides) Height measurement: Pitch angle of area scan camera 11 pixel distance from the top and bottom edges of the product Combined with pixel equivalent With camera focal length The calculation formula is: Correcting the height measurement deviation caused by the pitch angle; Aperture measurement: Extract the sub-pixel edge of the aperture, fit it to a circle, and calculate the diameter of the circle, which is the aperture. The fitted circle equation is: (in The sub-pixel coordinates of the center of the circle. (pixel radius of the circle), actual aperture ; Hole spacing measurement: Extracting the sub-pixel coordinates of the center of two holes. Calculate the straight-line distance between two points, which is the hole spacing. The calculation formula is: Contour measurement: Extract the sub-pixel coordinates of the actual product contour, compare them with the preset standard contour coordinates, and calculate the maximum deviation between the actual contour and the standard contour, which is the contour degree. ; Flatness measurement: Extract the sub-pixel coordinates of multiple feature points on the product surface, fit them to a plane, and calculate the maximum vertical distance from each feature point to this plane, which is the flatness. .

[0029] Dimensional acceptance judgment: The processing unit will process the actual measured values ​​of each critical dimension. , and the preset standard size , upper deviation , lower deviation (These are all preset parameters, which can be modified through the HMI interface on the control panel 8 to adapt to different product specifications.) A comparison and judgment are performed one by one, with the following logic: If the conditions are met... If the dimension is within acceptable limits, then the dimension is considered acceptable; if If the dimension is too small, it is judged as an out-of-tolerance "size too small" type, and the non-conformance type is marked as "size insufficient"; if If the dimension is found to be "too large", it is marked as "excessive". After all critical dimensions have been judged, the judgment result, actual measurement value and deviation value of each dimension are recorded. When the defect detection sub-model does not detect any appearance defects (i.e., the confidence level of all detected targets is <0.5, and it is judged as having no valid defects), and the judgment result of all key dimensions is "qualified", the product is judged as "qualified". The product is deemed "unqualified" when any of the following conditions are met: ① the defect detection sub-model detects at least one valid appearance defect (confidence level ≥ 0.5); ② at least one critical dimension is out of tolerance. Finally, the relevant qualified and unqualified data are classified, packaged, and sent to the main control system of the production line through the communication unit.

[0030] The above are merely embodiments of the present invention. The circuits, electronic components, and modules involved are all prior art, fully achievable by those skilled in the art, and require no further explanation. The scope of protection in this application does not involve improvements to the software and methods. Commonly known structures and characteristics in the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all prior art in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent.

Claims

1. A mobile visual quality inspection device for production lines, characterized in that: The system includes symmetrically arranged frames (1), with conveyor belts (2) between the frames (1). The upper and lower parts of the frames (1) are respectively connected to an upper guide rail (4) and a lower guide rail (20). Both the upper guide rail (4) and the lower guide rail (20) are provided with a first wheel rail (24). An AGV moving base plate (5) is provided between the upper guide rail (4) and the lower guide rail (20). The AGV moving base plate (5) is provided with a horizontal pulley (16). The horizontal pulley (16) is slidably connected to the first wheel rail (24). One end of the AGV moving base plate (5) is connected to a connecting column (15). The connecting column (15) is connected to a first connecting plate (6). The upper end of the first connecting plate (6) is connected to a first mounting plate (7). The upper end of the first mounting plate (7) is connected to a control plate (8). The lower end of the first mounting plate (7) is provided with a data acquisition component. The system also includes an auxiliary fixing component connected to the AGV moving base plate (5).

2. The mobile visual quality inspection device for production lines as described in claim 1, characterized in that: The acquisition component includes an area array camera (11), a first electric push rod (9) is connected to the upper end of the first mounting plate (7), the output shaft of the first electric push rod (9) is connected to a second mounting plate (10), and the area array camera (11) is connected to the second mounting plate (10).

3. A mobile visual quality inspection device for production lines as described in claim 2, characterized in that: The second mounting plate (10) is connected to sliding rods (27) at both ends. The first mounting plate (7) is provided with through holes. The sliding rods (27) are slidably connected to the through holes. The upper end of the sliding rods (27) is connected to a connecting block (28). The width of the connecting block (28) is greater than the width of the through holes.

4. A mobile visual quality inspection device for production lines as described in claim 3, characterized in that: The auxiliary fixing component includes a friction block (14), the AGV moving base plate (5) is connected to a second connecting plate (12), the second connecting plate (12) is connected to a second electric push rod (13), and the output shaft of the second electric push rod (13) is connected to the friction block (14).

5. A mobile visual quality inspection device for production lines as described in claim 4, characterized in that: The frame (1) is connected to a second fixed plate (21) on the side near the AGV moving base plate (5). The second fixed plate (21) is connected to a pressure sensor (22). The pressure sensor (22) is provided with a friction surface layer (29) at the end away from the second fixed plate (21). The friction surface layer (29) cooperates with the friction block (14).

6. A mobile visual quality inspection device for production lines as described in claim 5, characterized in that: The lower end of the AGV mobile base plate (5) is connected to a first fixed plate (18), and the lower end of the first fixed plate (18) is connected to a vertical pulley (19). The inner side of the lower guide rail (20) is provided with a second wheel rail (26), and the vertical pulley (19) is slidably connected to the second wheel rail (26).

7. A mobile visual quality inspection device for production lines as described in claim 6, characterized in that: The upper end of the AGV mobile base plate (5) is connected to a slider (17), and the inner side of the upper guide rail (4) is provided with a sliding groove (23). The slider (17) and the sliding groove (23) are slidably connected.

8. A mobile visual quality inspection device for production lines as described in claim 7, characterized in that: The friction surface layer (29) has a separator ball (25) on the side away from the pressure sensor (22), and the separator ball (25) is used to divide the pressure sensor (22) into sections.

9. A mobile assembly line visual quality inspection system, characterized in that: It includes an acquisition unit, a preprocessing unit, a processing unit, and a communication unit; The acquisition unit is used to acquire images of products on the production line; The preprocessing unit is used to perform deblurring preprocessing on the acquired images; The processing unit is used to extract features, identify defects, and measure dimensions of the preprocessed image, and output the quality inspection results in real time. The communication unit is used to transmit the quality inspection results.

10. A mobile assembly line visual quality inspection system as described in claim 9, characterized in that: The preprocessing unit includes an exposure duration control unit and an image compensation control unit. The exposure duration control unit is used to calculate the motion displacement of the product based on the operating speed of the production line, and generate an exposure duration control signal and image compensation parameters based on the motion displacement. The exposure duration control signal is used to control the exposure time of the high-speed area array camera to be less than the product motion time corresponding to a single pixel. The image compensation control unit is used to perform deblurring preprocessing on the acquired image based on the image compensation parameters.

Images