Method for detecting the state of a chassis support plate positioning pin

Abstract

The application belongs to the field of automatic detection of automobile manufacturing industry, and particularly relates to a detection method for the state of a positioning pin of an automobile chassis support plate, comprising the following steps: S1, acquiring local images of multiple automobile chassis support plates and adjusting and splicing the local images into a complete automobile chassis support plate image; S2, pre-processing the image collected by a camera; S3, acquiring a qualified detection positioning pin insertion and extraction state template and a support plate positioning template through an image training model; S4, using a PLC to receive a vehicle model information message, analyzing the message, importing the image shot by the camera, and detecting the state of the positioning pin by using a detection algorithm; S5, collecting and comparing the detection results of the positioning pins to determine whether all the positioning pins of the support plate are correct; if all the positioning pins are correct, the detection results are exported and the support plate is released; if the positioning pins are incorrect, an incorrect pin result is drawn to prompt manual modification. The application guarantees the accuracy of the detection of the state of the positioning pin and improves the detection efficiency of the entire chassis support plate.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of automatic detection in the automobile manufacturing industry, and particularly relates to a detection method for the state of a positioning pin of an automobile chassis pallet. BACKGROUND

[0002] In the automobile chassis assembly process, various parts need to be precisely positioned on the production pallet through positioning pins and supporting surfaces, and the "two-point-one-surface" positioning principle is usually adopted to ensure assembly stability and accuracy. With the development of automobile flexible production, co-line production of different vehicle models has become the norm. In order to accommodate different vehicle models, the chassis pallet is usually designed as an adjustable structure, and the state of the positioning pin is changed by plugging, flipping or sliding to adapt to different part layouts.

[0003] At present, the adjustment of the state of the positioning pin during model switching mostly relies on manual operation. Since there are many types of vehicles, the number, position and state of the positioning pins corresponding to each vehicle model are different, and the operator needs to check and adjust them one by one according to the production plan. This process not only requires a lot of work, but also is prone to problems such as omission, confusion or misplacement, resulting in inaccurate pallet positioning and causing part damage, assembly abnormalities or even the entire production line to stop. Some enterprises have tried to introduce board prompts and label markings to assist, but it is still difficult to fundamentally avoid human errors, and there is a lack of effective confirmation mechanism for the actual pin state.

[0004] Therefore, there is an urgent need for a positioning pin state automatic identification method that can quickly and accurately judge the state of each pin on the pallet based on visual detection technology, and compare it with the vehicle model information to assist manual model switching, improve operation accuracy, reduce assembly risks and ensure production rhythm and product quality. SUMMARY

[0005] The technical problem to be solved by the present application is that in the process of co-line production of multiple vehicle models, the state of the pallet positioning pin needs to be adjusted manually, which is prone to problems such as omission, confusion or error, resulting in assembly defects, part damage or production line stoppage.

[0006] To this end, the present application provides a detection method for the state of a positioning pin of an automobile chassis pallet, which on the one hand ensures the accuracy of the detection of the state of the positioning pin, and on the other hand improves the detection efficiency of the entire chassis pallet.

[0007] The technical scheme adopted by the present application to solve its technical problem is:

[0008] A detection method for the state of a positioning pin of an automobile chassis pallet, comprising,

[0009] S1, obtaining local images of multiple automobile chassis pallets and adjusting and splicing them into a complete automobile chassis pallet image;

[0010] S2, preprocessing the image captured by the camera;

[0011] S3. Use an image detection algorithm to identify the image and obtain a qualified positioning pin insertion and removal status template and a pallet positioning template to be detected through an image training model;

[0012] S4. Use the PLC to receive the vehicle model information message, parse the message, import the image captured by the camera, and use the detection algorithm to detect the status of the positioning pin;

[0013] S5. Summarize and compare the inspection results of each positioning pin to determine whether all the positioning pins of the pallet are correct; if all the positioning pins are correct, export the inspection results and release the pallet; if the positioning pins are incorrect, draw the error pin results and prompt manual modification.

[0014] Furthermore, in step S3, the basic information required by the image detection algorithm is first entered, and then the image template is trained based on the entered basic information. The image template training process includes: constructing multiple image scales with different resolutions, and calling the Sobel and Canny edge detection algorithms in parallel on each scale image, fusing the extraction results, and finally uniformly upsampling and weighted integration to obtain the final edge feature map; extracting features based on the fused edge feature map to train the template, and the qualified trained template is used as the target image template.

[0015] Furthermore, in step 3, each positioning pin has different position states, and images containing different positioning pin insertion and removal states need to be obtained respectively, and the region of interest containing the positioning pin insertion and removal states is framed from the image.

[0016] Furthermore, in step 3, during the template training process, the template is verified. When the error of image matching using the current template for multiple consecutive times is ≤ When the pixel size is 1, the template is judged to be qualified.

[0017] Furthermore, in step S4, based on the current pallet vehicle model information in the PLC message, it is determined whether the vehicle model exists in the detection system and whether the current vehicle model needs to be detected.

[0018] Furthermore, in step S4, the process of detecting the state of the positioning pin using the detection algorithm includes:

[0019] The pallet positioning template is used to perform template matching on the captured image. The actual position offset of the pallet is determined by calculating the best matching position of the template in the current image. The position locator is then generated to achieve the position reference alignment of the pallet in the image.

[0020] Bind each area to be detected to the above locator by relative coordinates, so that each detection frame can be adaptively adjusted according to the actual position and angle of the pallet;

[0021] All detection areas bound to the locator are inspected one by one. During the inspection process, the locating pin status detection algorithm is used to determine the current status of the locating pin based on the characteristic differences of different types of locating pins. The inspection results are compared with the standard status to draw a detection conclusion.

[0022] Furthermore, in the positioning pin state detection algorithm, for the positioning pin in the overturned state, based on the image feature differences between the normal upright state and the overturned state, a fixed rectangular area with obvious grayscale changes is selected as the detection area; the center coordinates of the detection area in the image I (x, y) are extracted. , width W f , height H f , and calculate the average gray value MGV of the area under different states f By comparing the gray value of the current image with the gray value of the standard template, the actual state of the tipping pin is determined, and the average gray value MGV f The calculation method is as follows: .

[0023] Furthermore, in the positioning pin state detection algorithm, for the moving pin, the detection area of ​​the moving pin is determined by comparing the change in grayscale distribution in the image before and after the movement; the center coordinates of the area are extracted. , width W f , height H f , and calculate the average gray value MGV of the area y , the actual position and status of the moving pin are determined based on the grayscale value difference.

[0024] Furthermore, in the positioning pin status detection algorithm, for the bolt insertion and extraction pins, it is first determined whether there is an obvious grayscale value change in the detection area. If the grayscale change is significant, the fixed ROI grayscale value comparison method is used for detection; if the grayscale change is not obvious, the edge template matching algorithm is used to extract edge features for identification.

[0025] Furthermore, in step S5, the detection results of each positioning pin are received, including the overturned pin status (F), the movable pin position and status (M), and the bolt insertion and extraction pin type and existence status (P), and whether the pallet release requirements are met is determined according to the following judgment formula: The release judgment formula is: , where the result of each pin status judgment is defined as a Boolean value: : Status of the i-th tipping pin (1 for normal, 0 for abnormal); : Status of the i-th moving pin (correct position and not missing); : Status of the i-th bolt insertion and extraction pin (correct type and existence), N is the number of locating pins that should exist on the pallet, R=1 means that the release condition is met, otherwise it is not met.

[0026] The beneficial effect of the present invention is that it improves the problem of easy errors and low efficiency in the prior art of relying on manual confirmation of pin status, reduces the risk of damage to parts, and reduces the risk of rework and line stoppage due to operational and detection errors by automatically matching visual inspection results with vehicle model information, thereby improving the rhythm stability and automation level of the vehicle assembly line.

[0027] This application establishes a stable and efficient information interaction mechanism between PLC and image detection algorithm, realizing functions such as synchronization with production line rhythm, accurate vehicle model identification and result feedback release. It uses vehicle model message as the core trigger basis and combines image recognition algorithm to intelligently detect the status of positioning pins on the pallet, thereby forming a closed-loop process of automatic release or prompt for re-inspection, thereby improving the degree of automation of pin status detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The present invention will be further described below with reference to the accompanying drawings and examples.

[0029] Figure 1 The present invention is a flow chart of a method for detecting the state of a positioning pin of an automobile chassis support plate.

[0030] Figure 2 It is a schematic diagram of the capture images of different cameras of the present invention.

[0031] Figure 3 is a schematic diagram of the tipping pin of the present invention.

[0032] Figure 4 It is a schematic diagram of the plug pin of the present invention.

[0033] Figure 5 It is a schematic diagram of an image template of a bolt insertion and extraction pin of the present invention.

[0034] Figure 6 Schematic diagram of the positioning image template of the support plate of the present invention.

[0035] Figure 7 It is a flow chart of the feature extraction method based on multi-scale edge fusion in the present invention.

[0036] Figure 8 It is a flow chart of communication interaction between the detection system of the present invention and the PLC system.

[0037] Figure 9 Schematic diagram of a camera image with a detection frame according to the present invention. DETAILED DESCRIPTION

[0038] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0039] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying 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, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, features defined as "first" or "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0040] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0041] Reference Figure 1 A method for detecting the state of a positioning pin of a chassis support plate of an automobile comprises the following steps:

[0042] S1. Set up the camera and light source to the appropriate position, adjust the imaging parameters, and ensure the image is clear and stable;

[0043] Due to the limitations of the site height, camera field of view and pixel accuracy, one camera cannot capture the entire chassis pallet within the frame. Therefore, four cameras are required to capture different areas to obtain a complete pallet image. The image must also be adjusted to ensure clarity. Figure 2 As shown, the images at different locations are captured by four cameras.

[0044] Specifically,

[0045] S11. Select four 2448×2048 resolution, 5-megapixel area scan cameras and pair them with four 800mm white bar lights. Mount the cameras and lights 1.75 meters above the chassis tray, pointing downward. The spacing between the cameras is 750mm horizontally and 790mm vertically, resulting in a field of view of 924×773mm for each camera. Install the lights inside the four cameras.

[0046] S12. Because the pallet is large and the pins are scattered, a single camera cannot cover the entire pallet. Therefore, four cameras are used to simultaneously capture the upper left, lower left, upper right, and lower right areas of the pallet. Overall inspection is achieved through image stitching. The camera's image resolution parameters are adjusted, and the image of non-product areas is cropped.

[0047] S13. According to the actual production pallet, the imaging picture is flipped horizontally or vertically to ensure that the picture sequence of the four cameras is correct and the production pallet picture is fully displayed;

[0048] S14. Adjust the camera exposure so that the average grayscale of all positioning pin areas in the production tray reaches the target grayscale, ensuring that the outlines are clearly visible. Exposure is performed by adjusting the camera aperture, exposure time, and gain parameters to ensure clear and stable imaging.

[0049] S2. Preprocess the image captured by the camera to ensure that the main position features of the image are clear, and create a bolt insertion and extraction pin template and a pallet positioning template. The steps of preprocessing the image in S2 are as follows:

[0050] S21, converting the image into a grayscale image to facilitate subsequent image processing;

[0051] S22, using median filtering to remove salt and pepper noise in the image;

[0052] S23. Use the Canny operator to sharpen the image, compensate for the contour, enhance the edge and grayscale jump parts, thereby improving the image clarity;

[0053] S24, enhancing the grayscale image to increase the contrast of the image;

[0054] S25, performing histogram equalization, by counting the frequency of occurrence of each gray value and equalizing the histogram, thereby improving the distribution of pixel gray values ​​and enhancing image brightness;

[0055] S26. Perform grayscale transformation, calculate the maximum and minimum grayscale values ​​of the image, and map them to the range of 0-255 to enhance the overall visual effect.

[0056] S27. Set the salient area in the image as the region of interest, apply the Canny operator and the Sobel gradient operator to it to extract edge information, and retain the contour, position and angle features.

[0057] Figure 3 The pin detection diagram of the tipping pin in the tipping state and the tipping state (the region of interest is set to the tipping pin axis position, Figure 3 The enlarged image at point A in the middle is a schematic diagram of the tipping pin in the fallen state. The tipping pin shaft position cannot detect the white area at the top of the tipping pin. Figure 3 The enlarged image at point B in the middle is a schematic diagram of the tipping pin in the flipped-up state. The white area at the top of the tipping pin is located directly above the pin shaft position). Figure 4 The detection diagram of the pin status in the two states of plugging and unplugging (the area of ​​interest is set to the bottom connection axis position, refer to Figure 4 Part A in the figure is a schematic diagram of the area of ​​interest when the pin is not installed. The bottom connecting shaft is exposed and the image display area is black. Figure 4 Part B in the figure is a schematic diagram of the region of interest with the pin installed. The bottom connecting shaft is blocked by the pin and is displayed in white. To ensure that each type of positioning pin can still be clearly imaged under different lighting conditions, image preprocessing is performed on the images captured by the camera in this embodiment. Figure 3 、 4 These are image samples of the tipping pin and plug-in pin after exposure adjustment. The positioning pin structure in the image is clearly visible, meeting the detection requirements.

[0058] S3. Use the image detection algorithm to identify the image, and obtain the qualified positioning pin insertion and removal status template and pallet positioning template to be detected through the image training model.

[0059] S31, enter the basic information required by the image detection algorithm, including the production pallet positioning information, all the positioning pins on the pallet to be detected, and the detection vehicle model information:

[0060] Specifically, the production pallet positioning information includes the pallet's horizontal pixel coordinate X0, vertical pixel coordinate Y0, and tilt angle R0 (rotation angle in the XY plane) in the image;

[0061] Information on all locating pins on the pallet that need to be inspected, including the number of locating pins on the pallet to be inspected, the type of each locating pin (such as tipping pin, bolt insertion pin, movable pin, etc.), and the coordinates of the inspection area corresponding to each locating pin;

[0062] The information of the vehicle model to be tested includes the vehicle model name, vehicle model number, the locating pins involved in the vehicle model, and the correct status of each locating pin.

[0063] S32, training image template based on the input basic information:

[0064] In this embodiment, if Figure 5 As shown, the image template of the bolt insertion pin, Figure 5 In the diagram, (a) is a schematic diagram with pin 1 installed, (b) is a schematic diagram with pin 2 installed, and (c) is a schematic diagram without pins installed; Figure 6 Shown is an image template of the positioning mark of the pallet.

[0065] S321, the bolt insertion and extraction pin template needs to be created separately. The bolt insertion and extraction pin has three states: pin 1 state, pin 2 state, and no pin state. The training process of the bolt insertion and extraction pin image template is as follows:

[0066] An area array camera is used to collect image samples containing the three states mentioned above. The region of interest (ROI) containing the bolt insertion and extraction pins is calibrated in each image. To extract more stable and complete edge features in the target image, a feature extraction method based on multi-scale edge fusion is proposed:

[0067] This method takes the original image as input, constructs multiple image scales with different resolutions, and calls two classic edge detection algorithms, Sobel and Canny, in parallel on each scale image. The extracted results are fused and finally uniformly upsampled and weighted integrated to obtain the final edge feature map.

[0068] like Figure 7 As shown in the figure, for the extracted image of the region of interest, the original image is resized to generate a set of reduced images (such as 0.5x, 0.25x, etc.) to form a multi-scale image pyramid, and obtain images at several scale levels: , the sizes of images at different scales are inconsistent, and image features can be observed from different resolutions: large-scale images (low resolution) focus more on the overall outline; small-scale images (high resolution) can retain more detailed textures.

[0069] For each scale image, Sobel operator and Canny operator are executed in parallel in different threads to extract edge information, and the edge response map at the current scale is obtained by weighted fusion. Specifically, the k-th scale image The edge fusion feature extraction method is as follows:

[0070] ,

[0071] in, For the Layer-scale images, 、 is the edge fusion weight coefficient, satisfying The fusion weights can be adjusted according to different task scenarios, providing good flexibility and adjustability. The Sobel operator is sensitive in gradient direction detection and can extract detailed edge information; the Canny operator performs well in noise resistance and edge connectivity, and is suitable for complex backgrounds or blurred images.

[0072] At each image scale, Sobel and Canny calculations are asynchronously executed by separate threads, enabling rapid parallelization and improving overall execution speed. The isolation of the two operators within a thread facilitates modular invocation and scalable deployment. The independent output results of different threads facilitate parameter tuning, debugging, and fault location for individual operators, and also promotes efficient CPU utilization in hardware deployment. The output results of each thread are fused in a weighted manner, preserving both detail and overall structure at the current scale, resulting in stronger edge representation capabilities.

[0073] To ensure the consistency of the fusion image size, the fusion results of each scale are After upsampling to the original size, multi-scale weighted integration is performed to obtain a complete edge feature map:

[0074] ,

[0075] in, Fusion weights for different scales, satisfying Multi-scale weighted integration aligns edge results extracted at different scales in a unified space, ensuring that the final fused image has the same resolution as the original image, facilitating subsequent processing. This integration of cross-scale information improves the stability and integrity of edge detection, particularly in industrial tasks such as bolt inspection. It effectively balances details with the overall outline, improving detection stability and accuracy.

[0076] Finally, based on the fused edge feature map, the contour, position, and angle features of the bolt insertion and extraction pins are extracted and fed into the template training module as training samples to generate qualified image templates. The generated bolt insertion and extraction pin templates cover three pin states and can be used for automatic state discrimination in subsequent detection and recognition tasks.

[0077] S322, the training process of the pallet positioning template includes:

[0078] An area array camera is used to capture the image of the pallet, and the region of interest containing the main features of the pallet is framed from the image. The edge information of the pallet positioning area is extracted from the region of interest using the edge fusion feature extraction method, and the contour features, position features, and angle features of the pallet positioning area are retained. The template is trained using the region of interest and its features, and the qualified template is used as the pallet positioning template.

[0079] It should be noted that the bolt insertion and extraction pin has three different states, and three different template images are used to train the template using the region of interest and its features. The qualified template is used as the template of the corresponding bolt insertion and extraction pin. For example, during template training, the template can be verified. When the error of image matching using the current template for 5 consecutive times is ≤ 1 pixel, it is determined that the template is qualified.

[0080] S4, using PLC to receive vehicle information message, analyzing the message, importing the image shot by the camera, and using detection algorithm to detect the pin state;

[0081] In this embodiment, after completing image acquisition and template configuration, and entering basic detection information, the system enters the detection execution phase. In order to realize the functions of synchronization with production line rhythm, accurate vehicle identification and result feedback release, the present application establishes a stable and efficient information interaction mechanism between the detection system and the PLC system. The interactive process takes the vehicle message as the core trigger, combines the image recognition algorithm to intelligently detect the pin state on the tray, and then forms a closed loop process of automatic release or prompt re-inspection.

[0082] Specifically, as shown in Figure 8 , it is the communication and interaction process diagram of the detection system and the PLC system; as shown in Figure 8 , it is the message structure of the PLC sending vehicle information, the length of which is fixed at 64 bits, and the 47th to 51st bits are used to identify the current vehicle information. As shown in Figure 9 , it is the position of all detection boxes in the image (as shown by the red detection box in the figure). After the system receives the message, it enters the image acquisition and recognition process, and feeds back OK or NG instructions according to the recognition result, to realize the control logic of whether to release the tray.

[0083] On this basis, step S4 specifically includes the following sub-steps:

[0084] S41, after the tray reaches the designated detection station, the sensor is triggered, the PLC message information is sent, and the detection system receives the message and enters the working process. The message is analyzed to obtain the current tray vehicle information, and the existing vehicle information is searched to determine whether the current vehicle needs to be detected and whether the detection system has the vehicle information.

[0085] S42, the specific content of the PLC message format is: the length of the message is fixed at 64 bits; the useful information in the message that needs to be analyzed is the 47th-51st bits, which are 5-bit vehicle information; different vehicle information represents different vehicle models.

[0086] S43. If the vehicle model information exists normally in the detection system, the detection system will take pictures and perform detection, and judge whether the result is correct based on the feedback from the detection system. If it is correct, an OK message will be sent and the pallet will be released; if it is wrong, the workflow ends and waits for the PLC to send a re-inspection message or a new vehicle model information message.

[0087] Furthermore, the state detection algorithm includes the following steps:

[0088] S44, performing a template matching operation on the captured image using the pallet positioning template (the template is the image template trained in step S3), calculating the best matching position of the template in the current image, determining the actual position offset of the pallet (including horizontal offset, vertical offset, and angular offset), and thereby generating a position locator (Locator) to achieve position reference alignment of the pallet in the image;

[0089] S45, bind each area to be detected (detection frame) to the pallet locator through relative coordinates, so that the position and angle of the detection frame can be dynamically adjusted according to the actual posture of the pallet. The specific steps include:

[0090] Using the template matching algorithm, the reference locator position and rotation angle of the pallet are accurately located in the captured image as the spatial reference of the pallet;

[0091] Based on the real-time coordinates of the locator in the image (including translation displacement and rotation angle), the spatial transformation matrix of each detection frame relative to the locator is calculated to achieve dynamic position and posture adjustment of the detection frame;

[0092] Through this transformation, all detection frames can be adaptively adjusted according to the actual position and angle of the pallet, ensuring that the detection area always accurately covers the target area;

[0093] Combining multi-scale image pyramid technology and multi-threaded parallel edge feature extraction, the efficiency and robustness of the locator calculation are improved, thereby enhancing the stability and accuracy of the overall detection system.

[0094] For example, in an image, we first use image processing methods (such as template matching or edge feature extraction) to find a representative "location area" or "feature point" on the pallet. The position and orientation of this area in the image represent the actual position and angle of the pallet.

[0095] We can obtain the pallet's rotation angle by fitting a straight line to the pallet's edge, and determine its coordinates in the image using the pallet's center point. This ultimately creates a "locator" consisting of a position (x, y) and an angle (r), which serves as a reference for subsequent transformations.

[0096] Each detection frame in the system is designed to be independently set relative to the locator, that is, the coordinates and angles of each detection frame are "referenced to the pallet".

[0097] The locator in the pallet positioning template is set to (x0, y0, r0). When the new locator of the actual pallet position is (x1, y1, r1), we only need to calculate the new transformation relationship (x0-x1, y0-y1, r0-r1) based on the new locator value and apply this transformation to the actual relative coordinates of each detection frame.

[0098] In this way, each detection frame will move with the pallet. No matter how the pallet moves or rotates, the detection frame will always follow accurately, ensuring that the target detection area is always in the correct position, thereby improving overall recognition accuracy and stability.

[0099] S46. Perform inspection operations on all detection areas bound to the locator one by one. During the inspection process, based on the feature differences of different types of locating pins, the corresponding locating pin state detection algorithm (including grayscale difference detection, edge feature extraction, template matching, etc.) is used to determine the current state of the locating pin. The inspection result is compared with the standard state to draw a detection conclusion.

[0100] Furthermore, the state detection algorithm of the positioning pin includes:

[0101] S461. For the overturned pin, based on the image feature differences between the pin in a normal upright state and in an overturned state, a fixed rectangular area with obvious grayscale changes is selected as the detection area;

[0102] For image I(x, y), extract the center coordinates of the detection area , width W f , height H f , and calculate the average gray value MGV of the area under different states f ,By comparing the difference between the gray value of the current image and the gray value of the standard template, the actual state of the tipping pin is determined.

[0103] Mean gray value MGV f The calculation method is as follows:

[0104] .

[0105] S462, determining a detection area for the moving pin by comparing changes in grayscale distribution in the image before and after the moving pin;

[0106] Extract the center coordinates of the area , width W f , height H f, and calculate the average gray value MGV of the area y , the actual position and status of the moving pin are determined based on the grayscale value difference.

[0107] S463: For bolt insertion and extraction pins, first determine whether there is a significant grayscale value change in the detection area. If the grayscale change is significant, a fixed ROI grayscale value comparison method is used for detection. If the grayscale change is not significant, an edge template matching algorithm is used to extract edge features for identification. By matching with three standard bolt insertion and extraction pin templates (i.e., "Pin 1," "Pin 2," and "No Pin"), the type of the current pin and its presence status are identified.

[0108] S5. Summarize and compare the inspection results of each positioning pin to determine whether all the positioning pins of the pallet are correct; if all the positioning pins are correct, export the inspection results and release the pallet; if the positioning pins are incorrect, draw the error pin results and prompt manual modification.

[0109] In this embodiment, after completing image acquisition and positioning pin status recognition, the system needs to make a comprehensive judgment on the recognition results to determine whether the current pallet meets the release conditions.

[0110] Step S5 includes the following steps:

[0111] S51. Receive the inspection results of each positioning pin, including the tipping pin status (F), the moving pin position and status (M), and the bolt insertion and extraction pin type and presence status (P). Determine whether the pallet release requirements are met according to the following judgment formula:

[0112] Define the status judgment result of each pin as a Boolean value:

[0113] : Status of the i-th tipping pin (1 for normal, 0 for abnormal);

[0114] : Status of the i-th moving pin (correct position and not missing);

[0115] : The status of the i-th bolt insertion and extraction pin (correct type and exists).

[0116] The release decision formula is: ,

[0117] Where N is the number of positioning pins that should exist on the pallet, R=1 means that the release condition is met, otherwise it is not met.

[0118] S52. If all the positioning pins are in the correct state, a test result file is generated and the pallet is released.

[0119] S53. If any positioning pin is in an abnormal state, the abnormal pin is marked, an error mark image is automatically drawn, and the system prompts manual correction or processing until the fault is eliminated.

[0120] This embodiment improves the detection method to significantly improve the accuracy of the detection of the positioning pin state and the detection efficiency of the entire chassis support plate. The advantages of this method compared with the existing method are shown in Table 1.

[0121] Table 1 Comparison of detection methods

[0122]

[0123] In summary, the method for detecting the state of the positioning pins of the automobile chassis support plate of the present invention has the following advantages:

[0124] In terms of accuracy: In actual testing of a certain model of production line, the visual recognition method of the present invention can achieve an accuracy of 99.82% in multiple consecutive rounds of detection, verifying its recognition stability and practicality under complex working conditions.

[0125] Identification Efficiency: Based on field measurements, this system can inspect over 600 chassis pallets daily, with an average single-pallet inspection time of less than one second. This represents an improvement of over 30 times in efficiency compared to manual methods. By utilizing a parallel processing mechanism combining image preprocessing, adaptive feature region positioning, and a state recognition algorithm, this system significantly shortens the inspection cycle while ensuring high recognition accuracy. This system offers excellent field adaptability and batch inspection capabilities, meeting the high-paced, high-consistency inspection requirements of vehicle assembly lines.

[0126] The present invention improves the error-prone and inefficient problem of relying on manual pin status confirmation in the prior art, reduces the risk of damage to parts, and reduces the risk of rework and line stoppage due to operational and detection errors by automatically matching visual inspection results with vehicle model information, thereby improving the rhythm stability and automation level of the vehicle assembly line.

[0127] With the above-described preferred embodiments of the present invention as a guide, and with reference to the above description, relevant personnel are fully capable of making various changes and modifications without departing from the technical spirit of this invention. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for detecting the state of a positioning pin of a chassis support plate of an automobile, characterized in that: The following steps are included: S1, obtaining multiple partial images of the automobile chassis pallet and adjusting and splicing them into a complete automobile chassis pallet image; S2, preprocessing the image captured by the camera; S3. Use an image detection algorithm to identify the image and obtain a qualified positioning pin insertion and removal status template and a pallet positioning template to be detected through an image training model; S4. Use the PLC to receive the vehicle model information message, parse the message, import the image captured by the camera, and use the detection algorithm to detect the status of the positioning pin; In step S4, the process of using the detection algorithm to detect the state of the positioning pin includes: The pallet positioning template is used to perform template matching on the captured image. The actual position offset of the pallet is determined by calculating the best matching position of the template in the current image. The position locator is then generated to achieve the position reference alignment of the pallet in the image. Bind each area to be detected to the above-mentioned locator by means of relative coordinates, so that each detection frame can be adaptively adjusted according to the actual position and angle of the pallet; All detection areas bound to the locator are inspected one by one. During the inspection process, the locating pin status detection algorithm is used to determine the current locating pin status based on the characteristic differences of different types of locating pins. The inspection results are compared with the standard status to draw a detection conclusion. S5. Summarize and compare the inspection results of each positioning pin to determine whether all the positioning pins of the pallet are correct; if all the positioning pins are correct, export the inspection results and release the pallet; if the positioning pins are incorrect, draw the error pin results and prompt manual modification.

2. The method for detecting the state of the positioning pin of the automobile chassis support plate according to claim 1, characterized in that: In step S3, the basic information required by the image detection algorithm is first entered, and then the image template is trained based on the entered basic information. The image template training process includes: constructing multiple image scales with different resolutions, and calling the Sobel and Canny edge detection algorithms in parallel on each scale image, fusing the extraction results, and finally uniformly upsampling and weighted integration to obtain the final edge feature map; extracting features based on the fused edge feature map to train the template, and the qualified trained template is used as the target image template.

3. The method for detecting the state of the positioning pin of the automobile chassis support plate according to claim 2, characterized in that: In step S3, each positioning pin has different position states, and images containing different positioning pin insertion and removal states need to be obtained respectively, and the region of interest containing the positioning pin insertion and removal states is framed from the image.

4. The method for detecting the state of the positioning pin of the automobile chassis support plate according to claim 3, characterized in that: In step S3, during the training process of the image template, the image template is verified. When the error of image matching using the current image template for multiple consecutive times is less than When the pixel value is 1, the image template is judged to be qualified for training.

5. The method for detecting the state of the positioning pin of the automobile chassis support plate according to claim 1, characterized in that: In step S4, based on the current pallet model information in the PLC message, it is determined whether the model exists in the detection system and whether the current model needs to be detected.

6. The method for detecting the state of the positioning pin of the automobile chassis support plate according to claim 1, characterized in that: In the state detection algorithm of the positioning pin, for the overturned pin, based on the image feature difference between the normal upright state and the overturned state, a fixed rectangular area with obvious grayscale change is selected as the detection area; the center coordinates of the detection area in the image I (x, y) are extracted. , width W f , height H f , and calculate the average gray value MGV of the area under different states f By comparing the gray value of the current image with the gray value of the standard template, the actual state of the tipping pin is determined, and the average gray value MGV f The calculation method is as follows: .

7. The method for detecting the state of the positioning pin of the automobile chassis support plate according to claim 1, characterized in that: In the state detection algorithm of the positioning pin, for the moving pin, the detection area of ​​the moving pin is determined by comparing the changes in the grayscale distribution in the image before and after the movement; the center coordinates of the area are extracted. , width W f , height H f , and calculate the average gray value MGV of the area y , the actual position and status of the moving pin are determined based on the grayscale value difference.

8. The method for detecting the state of the positioning pins of the automobile chassis support plate according to claim 1, characterized in that: In the state detection algorithm of the positioning pin, for the bolt insertion and extraction pin, it is first determined whether there is an obvious grayscale value change in the detection area. If the grayscale change is significant, the fixed ROI grayscale value comparison method is used for detection; if the grayscale change is not obvious, the edge template matching algorithm is used to extract edge features for identification.

9. The method for detecting the state of the positioning pin of the automobile chassis support plate according to claim 1, characterized in that: In step S5, the detection results of each positioning pin are received, including the overturned pin state F, the movable pin position and state M, and the bolt insertion and extraction pin type and existence state P. The following judgment formula is used to determine whether the pallet release requirements are met: The release judgment formula is: , where the result of each pin status judgment is defined as a Boolean value: : The state of the i-th tipping pin, 1 means normal, 0 means abnormal, i=1,2,..., ; : Status of the i-th moving pin, 1 means the position is correct and not missing, i=1,2,..., ; : Status of the i-th bolt insertion and extraction pin, 1 means the type is correct and exists, i=1,2,..., , R=1 means the release condition is met, otherwise it is not met.

Images