End-cloud collaborative automobile fastener quality data traceability system

Abstract

This invention relates to the field of product quality traceability technology, including an edge-cloud collaborative automotive fastener quality data traceability system. The system comprises an edge offset parsing module, a traceability tag binding module, a sequence trend construction module, a channel path identification module, and a collaborative traceability aggregation module. In this invention, by directly acquiring and parsing fastener image boundary information, offset state tags are established and image state sets are formed. These tags are then bound to device batches based on time information. Clustering of change frequency and span patterns in continuous frames is performed, and evolution paths are constructed by combining confirmed abnormal trajectories from historical detection records. These paths are then matched and identified with current trends to complete the screening and classification of typical offset evolution patterns. Synchronous cloud display enables edge-cloud collaborative data integration and dynamic traceability, significantly enhancing the ability to identify quality anomalies, provide trend warnings, and locate paths. This effectively solves the problems of insufficient perception and low traceability efficiency in existing systems.

Description

Technical Field

[0001] This invention relates to the field of product quality traceability technology, and in particular to a mobile terminal cloud collaborative automotive fastener quality data traceability system. Background Technology

[0002] Product quality traceability technology involves the recording, storage, management, and retrieval of quality information at each stage of a product's entire lifecycle. This includes the collection, coding, information storage, correlation tracking, and visualization management of quality data. By constructing a complete data chain from raw materials, production, testing, logistics to sales, it supports enterprises in locating the source of product quality problems and tracing responsibility. It has significant application value in manufacturing, especially in industries involving the precision assembly of parts. Among these, the traditional automotive fastener quality data traceability system refers to a system used to record, manage, and trace quality-related data generated during the manufacturing and assembly process of automotive fasteners. It uses QR codes, barcodes, or RFID tags as unique identifiers on fasteners or their packaging. Handheld scanners or industrial cameras are used to complete coding identification and data uploading at each production stage. The backend database stores and correlates information such as production batch, torque value, tightening angle, operation time, equipment number, and operator. When quality problems occur, the system uses the identifier to trace back the corresponding process and data source, assisting in defect analysis and responsibility determination.

[0003] Traditional automotive fastener quality data traceability systems rely on QR codes, barcodes, or RFID tags for information identification. Data collection at each stage requires manual or semi-automatic operation using scanners or industrial cameras. This system suffers from structural limitations due to its dependence on external tag media. Image content is not directly collected and analyzed, resulting in a lack of perception of changes in the actual state of parts. Data can only be traced back after the fact based on the tag information, making it difficult to identify quality problems early and judge their evolution trends. In addition, inconsistent information upload rhythms at each stage lead to insufficient perception of potential correlations between various types of quality data, limiting the ability to penetrate and analyze the quality problem chain. This makes it difficult to achieve efficient traceability and rapid location in complex manufacturing scenarios, affecting the overall traceability process's response efficiency and traceability depth. Summary of the Invention

[0004] To address the technical problems existing in the prior art, embodiments of the present invention provide an edge-cloud collaborative automotive fastener quality data traceability system. The technical solution is as follows: On the one hand, it provides an edge-cloud collaborative automotive fastener quality data traceability system, which includes: The edge offset analysis module acquires images of automotive fasteners, reads the pixel coordinates of the bolt outline boundary and the template boundary in the image, calculates the pixel difference between the horizontal and vertical positions, filters the set of continuous boundary points, marks the abnormal offset state images, and outputs the offset state image set. Based on the offset state image set, the traceability tag binding module extracts the timestamp, device number and batch of the abnormal offset state image, combines and encodes them and binds them to the image tag information, and pushes them to the cloud tagging interface to generate a time batch binding tag set. The sequence trend construction module performs clustering identification and sequential pattern organization on the continuous distribution of image offset direction according to the time batch and the tag set, filters out segments with obvious offset trend changes and performs clustering and classification processing to generate image offset trend segment sequence. The channel path recognition module extracts historical inspection records of automotive fasteners, analyzes and determines historical abnormal evolution channel paths, and matches and identifies the current segment's directional distribution by combining the image offset trend segment sequence with the current segment's directional distribution. It then filters evolution channels that match the offset change direction and have consistent trajectory distribution positions, generating a quality evolution channel path list. The collaborative traceability and aggregation module locates the process node number corresponding to each path according to the quality evolution channel path list, integrates them into a quality data traceability chain by batch according to the equipment distribution, and displays them synchronously on the end-to-cloud interface to obtain automotive fastener quality data traceability records.

[0005] As a further aspect of the present invention, the abnormal offset state image is specifically determined by comparing the differences in the set of continuous boundary points with a set boundary offset recognition threshold to determine whether there are more than 5 consecutive boundary points in each direction whose offsets in the same direction exceed the set boundary offset recognition threshold. If the corresponding images in both the horizontal and vertical directions are satisfied at the same time.

[0006] As a further aspect of the present invention, the segment with obvious shift trend change specifically refers to extracting the frequency of shift direction changes between adjacent frames, combining the number of shift direction changes within every 10 frames with the span value of shift direction changes, and identifying segments in consecutive frames that have more than 6 shift direction changes and span changes exceeding two shift direction categories.

[0007] As a further aspect of the present invention, the historical anomaly evolution channel path specifically involves extracting the image sequences marked in the batches confirmed by quality re-inspection from the historical detection records, and constructing a corresponding trajectory path based on the offset direction order of the images arranged on the time axis and the corresponding process node number.

[0008] As a further embodiment of the present invention, the offset state image set includes offset direction classification markers, abnormal state image indexes, and pixel difference distribution maps; the time batch binding tag set includes image time indexes, equipment station numbers, and abnormal batch codes; the image offset trend segment sequence includes offset direction change trajectories, change frequency statistical features, and trend pattern grouping categories; the quality evolution channel path list includes direction sequence paths, process node mappings, and image sequence index mappings; and the quality data traceability record includes manufacturing node marker chains, batch traceability mapping chains, and end-to-cloud display interface links.

[0009] As a further aspect of the present invention, the edge offset parsing module includes: The image boundary extraction submodule acquires images of automotive fasteners, extracts the pixel coordinates of the bolt contour boundary and the template boundary in the image, performs contour recognition on the bolt image, obtains the actual boundary point set in the image through coordinate mapping, and generates an image boundary coordinate set. The position deviation calculation submodule reads the horizontal and vertical pixel coordinate axes to construct a boundary comparison index set based on the image boundary coordinate set, calculates the pixel difference between the bolt outline boundary and the template boundary in the horizontal and vertical directions, and performs a filtering operation on the continuous boundary point region in the difference sequence to obtain a continuous direction deviation difference sequence. The offset anomaly identification submodule performs a continuity threshold judgment operation on the horizontal and vertical directions based on the continuous directional deviation difference sequence. It compares each value in the difference sequence with the set boundary offset identification threshold, marks the abnormal offset state image, and generates an offset state image set.

[0010] As a further aspect of the present invention, the traceability tag binding module includes: The image information extraction submodule, based on the offset state image set, reads the timestamp and device number by embedding metadata in the image file header, extracts the batch number information by matching the production batch record through the image path, and generates an image traceability information group; The label content generation submodule, based on the image traceability information group, concatenates the timestamp, device number, and batch number to construct a traceability label string in a unified format, writes it into the corresponding label of the image, and generates a time-batch combined label dataset. The cloud-based binding and push submodule encapsulates the unique image identifier and combined label into a standard data structure based on the time-batch combined label dataset, constructs a data upload request body, pushes data through the cloud tagging interface and receives binding response information, and generates a time-batch binding label set.

[0011] As a further aspect of the present invention, the sequence trend construction module includes: The orientation distribution recognition submodule extracts the offset orientation identifier corresponding to each frame of the image according to the time batch binding tag set, forms a continuous sequence of offset orientation information of adjacent image frames, arranges them in chronological order to form an orientation distribution chain, groups and counts the repetitive orientation segments, and generates an image orientation distribution coding sequence. The frequency feature extraction submodule, based on the image orientation distribution encoding sequence, counts the number of orientation changes by sliding window, calculates the orientation number span value within each window, records the segments with significant changes, outputs the start and end frame number indices of the segments, and generates a set of orientation change frequency indicators. The trend segment clustering submodule extracts image frame segments that continuously meet the conditions of direction change frequency and span based on the set of direction change frequency indicators. It performs temporal adjacency and direction encoding feature similarity matching on each segment, groups the related segments into the same group, and outputs the corresponding set of image frame numbers to generate an image offset trend segment sequence.

[0012] As a further aspect of the present invention, the channel path identification module includes: The abnormal trajectory extraction submodule obtains the image sequence of abnormal batches confirmed by quality re-inspection from the historical inspection records, extracts the timestamp, offset direction and corresponding process node number of each image, sorts them in ascending order by image timestamp, outputs the direction evolution order and the corresponding structure of process node index, and generates a set of historical abnormal trajectory paths. The trend path matching submodule compares the similarity between the trend segment direction change and the historical trajectory direction evolution order based on the historical abnormal trajectory path set and the image offset trend segment sequence. It also performs position matching based on the time window of the trend segment and the process node number in the historical trajectory, filters trajectories with consistent direction patterns and node positions, and generates a trend path matching result set. The evolution channel filtering submodule performs trajectory continuity judgment on all matching items based on the trend path matching result set, filters paths that meet the requirements of the same direction change sequence and continuous node number distribution, outputs the matching path number, associated frame sequence and process node sequence in a structured manner, and generates a quality evolution channel path list.

[0013] As a further aspect of the present invention, the collaborative traceability and aggregation module includes: The node path localization submodule, based on the quality evolution channel path list, reads the node number sequence and the batch number to which the channel belongs in each channel path, establishes a node sequence index structure according to the path number and time order, performs node position parsing operation on each path and identifies the channel mapping relationship, and generates a channel node position index table. The batch chain integration submodule extracts the production line to which the equipment belongs based on the channel node position index table and the corresponding equipment number mapped by the node number. It then connects the equipment that the path passes through in the same batch in chronological order and establishes corresponding structured chain information for different batches to generate a quality traceability chain structure set. The end-to-cloud synchronous display submodule encapsulates the quality traceability chain structure set into a standard JSON format data structure required by the display interface, uploads it to the cloud display platform by batch, generates a traceability visualization display page through data interface calls, and synchronously writes it to the local archive database to generate automotive fastener quality data traceability records.

[0014] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: By directly acquiring and parsing the boundary information of fastener images, offset status markers are established and image status sets are formed. Combined with time information and equipment batch binding image tags, a serialized offset direction change trend is constructed. By clustering the change frequency and span patterns in continuous frames, a discernible trend segment sequence is formed. Then, combined with the confirmed abnormal trajectories in historical detection records, an evolution path is constructed and matched with the current trend to complete the screening and classification of typical offset evolution patterns. Finally, the abnormal path is mapped to the flowchart according to the process node number. Through synchronous display in the cloud, end-to-cloud collaborative data integration and dynamic traceability are realized, significantly enhancing the ability to identify the evolution of quality anomalies, provide trend warnings, and locate paths. This effectively solves the problems of insufficient perception and low traceability efficiency in the original system. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the end-to-cloud collaborative automotive fastener quality data traceability system provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the system framework of the present invention; Figure 3 This is a flowchart of the edge offset analysis module of the present invention; Figure 4 This is a flowchart of the traceability tag binding module of the present invention; Figure 5 This is a flowchart of the sequence trend construction module of the present invention; Figure 6 This is a flowchart of the channel path identification module of the present invention; Figure 7 This is a flowchart of the collaborative traceability and aggregation module of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] This invention provides an edge-cloud collaborative automotive fastener quality data traceability system, such as... Figure 1-2 The diagram shown illustrates an edge-cloud collaborative automotive fastener quality data traceability system. The system includes: The edge offset analysis module acquires images of automotive fasteners through the camera module, reads the pixel coordinates of the bolt outline boundary and the template boundary in the image, calculates the pixel difference in the horizontal and vertical positions respectively, filters the set of continuous boundary points, compares each difference in the set with the set boundary offset recognition threshold, and determines whether there are more than 5 consecutive boundary points in each direction whose offset exceeds the set boundary offset recognition threshold in the same direction. If the conditions in both the horizontal and vertical directions are met, the corresponding image is marked as an abnormal offset state image, and the offset state image set is output. The traceability tag binding module extracts the timestamp, device number and batch of abnormal offset state images from the offset state image set, combines and encodes them and binds them to the image tag information, and pushes them to the cloud tagging interface to generate a time batch binding tag set. The sequence trend construction module binds the label set according to the time batch, performs cluster recognition and sequential pattern sorting on the continuous distribution of image offset direction, extracts the frequency of change between offset directions of adjacent frames, and combines the number of direction changes and the span of direction changes within every 10 frames to filter segments with more than 6 direction changes and more than two direction categories in consecutive frames. The corresponding segments are marked as segments with obvious offset trend changes and are clustered and classified to generate image offset trend segment sequence. The channel path recognition module extracts historical inspection records of automotive fasteners, analyzes and determines historical abnormal evolution channel paths (extracts image sequences marked in batches confirmed by quality re-inspection in historical inspection records, constructs corresponding trajectory paths based on the offset direction order of images arranged on the time axis and the process node number), and matches and identifies the image offset trend segment sequence with the directional distribution of the current segment, filters evolution channels with the same offset change direction and consistent trajectory distribution position, and generates a list of quality evolution channel paths; The collaborative traceability and aggregation module locates the process node number corresponding to each path based on the quality evolution channel path list, integrates them into a quality data traceability chain by batch according to the equipment distribution, and displays them synchronously on the end-to-cloud interface to obtain automotive fastener quality data traceability records.

[0023] The offset status image set includes offset direction classification markers, abnormal status image indexes, and pixel difference distribution maps. The time batch binding tag set includes image time indexes, equipment station numbers, and abnormal batch codes. The image offset trend segment sequence includes offset direction change trajectories, change frequency statistical features, and trend pattern grouping categories. The quality evolution channel path list includes direction sequence paths, process node mappings, and image sequence index mappings. The quality data traceability record includes manufacturing node marker chains, batch traceability mapping chains, and end-to-cloud display interface links.

[0024] Specifically, such as Figure 2 , 3 As shown, the edge offset parsing module includes: The image boundary extraction submodule acquires images of automotive fasteners through the camera module, extracts the pixel coordinates of the bolt contour boundary and the template boundary in the image, performs contour recognition on the bolt image, obtains the actual boundary point set in the image through coordinate mapping, and generates an image boundary coordinate set. In the operation of acquiring images of automotive fasteners using a camera module, an industrial-grade line scan camera mounted above the detection area first performs high-resolution image acquisition of the area where the fasteners are located. The camera frame rate is set to 60 frames per second, and the exposure time is controlled within 2ms to ensure that the image is free of motion blur. The image resolution is set to 2048×1080 pixels to cover the entire detection range. The image data is transmitted to the storage buffer in the edge computing unit through the image acquisition interface. In the step of extracting the pixel coordinates of the image contour boundary, the Canny edge detection operator is used to extract the boundary line segments of the bolt contour in the image, and morphological erosion is combined to remove image noise interference. The set of boundary coordinate points is represented in two-dimensional coordinate form, such as the bolt boundary point coordinates as follows: Where n represents the number of boundary points, the corresponding template boundary in the image is a standard image already entered in the standard database, and the set of boundary point coordinates in the standard image is extracted with sub-pixel precision, denoted as . Then, the two sets of boundary points are compared, and the image boundary points are matched one-to-one with the template boundary points according to the position matching rules. The matching rule uses the nearest neighbor Euclidean distance method for pairing. For example, the matching distance of the i-th point is calculated as follows:

[0025] When the distance value is less than the set matching tolerance ε = 3 pixels (used to cover the acceptable pixel spacing after the mechanical positioning offset upper limit and edge detection jitter are superimposed; when the equivalent size of a single pixel obtained by camera calibration becomes smaller or the positioning error of the assembly fixture increases, ε can be adjusted synchronously with the equivalent size of a single pixel and the upper limit of the positioning error within the range of 2 to 4 pixels, and the adjustment principle is to keep ε not exceeding five percent of the bolt contour radius to avoid pairing across adjacent contour segments), then the image point and the template point are considered to be successfully paired. For successfully paired points, their boundary coordinate pairs are recorded to form a boundary pairing mapping table, as shown in the example below: Table 1 Image Boundary Pairing Coordinate Table

[0026] As shown in Table 1, the boundary points in the acquired image can be matched one by one with the boundary points of the standard template, establishing a complete spatial mapping relationship. Then, the boundary pairing results are stored and processed, and output to the subsequent calculation submodule for calling, finally generating a set of image boundary coordinates.

[0027] The position deviation calculation submodule reads the horizontal and vertical pixel coordinate axes from the image boundary coordinate set to construct a boundary comparison index set, calculates the pixel difference between the bolt contour boundary and the template boundary in the horizontal and vertical directions, and performs a filtering operation on the continuous boundary point region in the difference sequence to obtain a continuous direction deviation difference sequence. Based on the image boundary coordinate set, read the pairing mapping table between image boundary points and template boundary points. By extracting corresponding coordinate point pairs in the horizontal (x-axis) and vertical (y-axis) directions, construct a set of difference sequences. Let the horizontal difference of the i-th pair of boundary points be:

[0028] The longitudinal difference is:

[0029] Perform the above difference operation on each pair of coordinate points to form a difference vector group. For a specific example, if the set of image boundary points is {(120, 240), (121, 241), (122, 242)}, and the set of template boundary points is {(122, 239), (123, 240), (124, 241)}, then the difference sequence is Δx = {-2, -2, -2}, Δy = {1, 1, 1}. After constructing the horizontal and vertical difference sequences, a continuity filtering operation is further performed on the difference sequences. This operation is performed by setting a continuity judgment window W = 5 to filter the difference sequences. The sliding window is traversed, and five consecutive difference data are taken each time. It is determined whether the values ​​are all positive or negative in the same direction. If the condition is met, the boundary points in the window are considered to have continuous offset characteristics. Then, the boundary point indices corresponding to all windows that meet the rule are added to the continuous point set. The offset direction is set to positive offset (value is positive) or negative offset (value is negative). The difference is grouped in this way. The filtered set is used to describe the region with structural continuity offset characteristics, and finally a continuous direction deviation difference sequence is generated.

[0030] The offset anomaly identification submodule is based on the continuous directional deviation difference sequence. It performs continuous threshold judgment operations for the horizontal and vertical directions respectively. It compares each value in the difference sequence with the set boundary offset identification threshold item by item, selects a set of 5 or more consecutive points whose deviation in the same direction exceeds the set boundary offset identification threshold, marks the images that meet both the horizontal and vertical judgment conditions, and generates an offset state image set. During the process of obtaining the continuous directional deviation difference sequence, each difference set is read. Anomaly detection is performed on the horizontal and vertical difference sets separately. This operation sets a boundary offset recognition threshold of δ=3 pixels and performs a sliding judgment operation on each difference sequence. Assuming the sliding window length is 5, if there are 5 consecutive data points within the window and their absolute values ​​all satisfy... If the condition is met, the window is recorded as a horizontal or vertical offset abnormal window. The horizontal and vertical offsets are judged independently. If there are abnormal windows in both the horizontal and vertical directions in the image, the image is considered to have a bidirectional offset state. The current image number, boundary point position index, and offset direction are recorded in the abnormal image database, and an image offset state label file is generated for subsequent processing. In practical applications, assuming Δx={4, 5, 6, 7, 5, 2} and Δy={0, 1, 4, 5, 6, 7}, the first 5 items in Δx meet the offset judgment condition, and the 3rd to 7th items in Δy meet the condition. Therefore, it is determined that the image has continuous offset anomalies in both the horizontal and vertical directions, and the image is finally classified as an anomaly, and an offset state image set is generated.

[0031] Specifically, such as Figure 2 , 4 As shown, the traceability tag binding module includes: The image information extraction submodule is based on the offset state image set. It reads the timestamp and device number by embedding metadata in the image file header, extracts the batch number information by matching the production batch record through the image path, and generates an image traceability information group. Based on the offset state image set, image data files are retrieved one by one, and image embedding information fields are extracted. The timestamp field is read by parsing the image file header structure. This timestamp field uses standard UTC format to represent the image generation time with second precision; for example, December 1, 2024, 08:30:15 can be represented as 20241201083015. This field has a fixed length of 14 bits. Simultaneously, the device number field is parsed from the EXIF ​​extension area of ​​the image file. This field records the acquisition device number, stored as a 6-character string, such as LGC001. Furthermore, the production batch number to which the image belongs is parsed from the batch directory hierarchy nested within the image file path. The batch number typically corresponds to the production plan number for that day, and is in the format of a 5-digit numeric string. For example, if the batch number for that day is 00321, the corresponding path format is / batch_00321 / image_xxxx.jpg. The system extracts the numeric string from this path using a regular expression parsing function as the batch number. Then, the above three fields are combined to form an image traceability information structure. The structure fields are in the order of timestamp, device number, and batch number. The source of field values ​​is fixed, the field length is clear, the field content structure is clear, and it has uniqueness verification capabilities. In large-scale image data management, it can be used to uniquely identify the source information and attribution attribute of each image. Sample data is shown in Table 2. Table 2 Example of Image Traceability Fields

[0032] As shown in Table 2, the system can extract three core pieces of information from each image as needed, and combine them with field format standardization rules to construct a dataset with a unified structure and clear fields, ultimately generating an image traceability information group.

[0033] The label content generation submodule, based on the image traceability information group, concatenates the timestamp, device number, and batch number to construct a traceability label string in a unified format, writes it into the corresponding label of the image, and generates a time-batch combined label dataset. After acquiring the image traceability information group, the timestamp field, device number field, and batch number field are read from the information structure of each group. According to the field format and encoding specifications, the three fields are sequentially concatenated to generate a 25-digit string code. The concatenated structure is: 6-digit device number + 14-digit timestamp + 5-digit batch number. During the concatenation process, the length of each field is validated. If it is insufficient, leading zeros are added for alignment. For example, device number LGC1 is padded to LGC001, and batch number 321 is padded to 00321, ensuring uniform length and consistency in the concatenation logic. The concatenated structure is LGC0012024120108301500321, where LGC001 is the device number, 20241201083015 is the image generation time, and 00321 is the... The system performs format validity checks on the spliced ​​results for the production batch number. The checks include whether the total string length is 25 characters, whether the character set conforms to the alphanumeric mixed standard, and whether the field segment positions are correct. If valid, the information is written to the corresponding image label information field. The label field uses a JSON format structure embedded in the image metadata, with the field key being `trace_code` and the value being the spliced ​​result. Once written, the image label structure update is complete, indicating that the image has traceability identification attributes. An example structure is as follows: `LGC0012024120108301500321`. The system generates a corresponding label structure for each image and stores it uniformly in the image label database for subsequent binding and uploading operations, ultimately generating a time-batch combined label dataset.

[0034] The cloud-bound push submodule combines the tag dataset according to the time batch, encapsulates the unique image identifier and combined tag into a standard data structure, constructs a data upload request body, pushes data through the cloud tagging interface and receives binding response information, and generates a time batch bound tag set. After obtaining the time-batch combined label dataset, the system constructs an upload structure for each image and its corresponding traceability label. The structure fields include the image's unique identifier `image_id` and the label content `trace_code`. The image's unique identifier is generated by combining the image filename and timestamp. For example, if the image filename is `img_0105.jpg` and the timestamp is `20241201083015`, then the `image_id` is `img_0105_20241201083015`. The system encapsulates this unique identifier and the corresponding label code into an upload request body. The request body structure is in JSON format, with key-value pairs of `{image_id: img_0105_20241201083015, trace_code: L`. The system constructs an HTTP POST request (GC0012024120108301500321), carrying the uploaded content and attaching an authentication token field (Authorization). The token value is a dynamic key issued by the deployment platform. This request is sent to the specified cloud tagging interface URL. This interface accepts JSON structure and returns a status code of 200 and binding success information after successful upload. The system parses the returned response content and writes the response status to the image binding status record table. The record items include image_id, trace_code, push timestamp, and return status. If the interface returns a status other than 200, the upload is marked as failed and transferred to the retry queue. Finally, the system counts the successfully pushed image entries and outputs the upload confirmation details, ultimately generating a time batch binding tag set.

[0035] Specifically, such as Figure 2 , 5 As shown, the sequence trend building module includes: The orientation distribution recognition submodule is based on the time batch binding tag set. It extracts the offset orientation identifier corresponding to each frame of the image in the order of image time, forms a continuous sequence of offset orientation information of adjacent image frames, arranges them in chronological order to form an orientation distribution chain, and performs group statistics on the repetitive orientation segments to generate an image orientation distribution coding sequence. Based on the time-batch bound tag set, all images are first sorted in ascending order by timestamp field to ensure the temporal continuity of the frame sequence. Then, the offset direction field of each image is extracted. This field comes from the recognition results of the previous module and is expressed using discrete directional coding values. For example, left offset is set to 1, right offset to 2, vertical offset to 3, horizontal offset to 4, and no offset is set to 0. The system constructs a two-dimensional sequence group by combining the image number and the offset code. Based on this, the system continuously splices the images in frame order to form a direction sequence. For example, if the image frame order is F1 to F10, the corresponding directions are... The code is {1, 1, 2, 2, 2, 3, 3, 1, 2, 2}. After reading the direction sequence, the system performs a distribution classification and recognition operation. By statistically analyzing segments with the same direction code in consecutive frames and recording their start and end positions and direction values ​​(e.g., {F1, F2} is direction 1, {F3, F4, F5} is direction 2, {F6, F7} is direction 3), the system completes the structured aggregation of the offset direction distribution. The system then constructs this aggregation structure into a structured record in sequence. The fields include the start frame, end frame, direction code, and number of frames for direction duration, as shown in the example below. Table 3 Image Offset Direction Aggregation Structure

[0036] As shown in Table 3, the system has completed the clustered coding process of the offset direction. This structure can realize the temporal expression of the directional evolution structure of the image sequence, and finally generate the image directional distribution coding sequence.

[0037] The frequency feature extraction submodule is based on the image orientation distribution coding sequence. It counts the number of orientation changes in every 10 frames of images as a sliding window unit, and calculates the orientation number span value within each window. When the number of orientation changes is greater than 6 and the orientation coding span is greater than 2 types of orientation numbers, the corresponding time window is recorded as the segment with obvious changes. The starting and ending frame number indices of the segment are output to generate a set of orientation change frequency indicators. After obtaining the image orientation distribution encoding sequence, the system performs a sliding traversal operation in 10-frame units. Within each window, it counts the number of orientation changes and the orientation span value. The number of orientation changes is defined as the total number of different orientation encoding values ​​between adjacent frames, and the orientation span value is defined as the number of different orientation encoding value categories within the window. In the example, if a window sequence is {1, 1, 2, 2, 3, 3, 1, 1, 2, 2}, then the number of orientation changes is 5, and the number of encoding categories is 3. The system records these two indicator values ​​for each window as structure fields and performs threshold filtering, setting the threshold for the number of orientation changes to 6. The span threshold is 2 categories. The filtering logic is as follows: if the number of directional changes within the window is greater than 6 and the number of directional encoding types is greater than 2, then the window is marked as an abnormal fluctuation segment, and the corresponding frame segment is recorded as a high-frequency change segment. The system also marks the start frame index and end frame index of the segment. If the window covers frame numbers F21 to F30 and meets the above conditions, then the output structure is {start_frame: F21, end_frame: F30, change_count: 7, span_value: 3}. Finally, all segment records that meet the conditions are summarized to generate a set of directional change frequency indicators.

[0038] The trend segment clustering submodule extracts image frame segments that continuously meet the conditions of direction change frequency and span based on the direction change frequency index set. It performs temporal adjacency and direction encoding feature similarity matching on each segment, groups the related segments into the same group, and outputs the corresponding image frame number set to generate an image offset trend segment sequence. Based on the frequency index set of directional changes, the system extracts all frame intervals that meet the conditions as candidate trend segments. It then performs clustering and grouping by comparing the temporal continuity and directional coding feature similarity between candidate segments. The temporal continuity criterion is that the frame interval between the start frame of an adjacent segment and the end frame of the previous segment does not exceed 5 frames. Directional feature similarity is determined by whether the dominant directional codes in two segments overlap. If the number of overlapping directional codes accounts for at least 0.5% of their respective coding sets, then the directional features are considered to overlap. For example, if segment A has a directional coding set of {1, 2, 3} and segment B has {2, 3, 4}, then the directional overlap is 2 / 3 and 2 / 3, satisfying the similarity requirement. The system assigns the same clustering label to adjacent and similar segments and generates a list of all image frame numbers within the cluster group. The clustering results are coded and identified, with each group generating a unified trend segment ID code, such as TRS001, TRS002, etc. Finally, all cluster groups and their corresponding frame sets are output, generating an image offset trend segment sequence.

[0039] Specifically, such as Figure 2 , 6 As shown, the channel path identification module includes: The abnormal trajectory extraction submodule obtains the image sequence of abnormal batches confirmed by quality re-inspection from the historical inspection records, extracts the timestamp, offset direction and corresponding process node number of each image, sorts them in ascending order by image timestamp, outputs the direction evolution order and the corresponding structure of process node index, and generates a set of historical abnormal trajectory paths. The system retrieves image sequences of abnormal batches confirmed through quality re-inspection from historical inspection records. First, it calls the batch index records marked as abnormal in the production quality traceability database, sequentially retrieving a list of image numbers belonging to that batch. Then, it calls the timestamp, offset direction identifier, and process node number fields corresponding to each image. The timestamp precision is in the second range, and the offset direction uses a unified direction code value (e.g., 1 for left offset, 2 for right offset, 3 for top offset, 4 for bottom offset, and 0 for no offset). The process node number uses a unique production line process station number; for example, node numbers 101 to 120 represent the inspection stations that the batch of images passed through sequentially. The system sorts all images in ascending order by the timestamp field, forming an image stream based on time evolution sequence, and then extracts... After sorting, the offset direction and process node number of each image are used as the main index to construct a mapping path between the direction sequence and the process node sequence. For each path, the start timestamp, end timestamp, direction change pattern, and node number sequence are recorded. In actual implementation, if an abnormal batch number is B0032, containing image sequences IMG0101 to IMG0120, with the corresponding direction sequence {1, 1, 2, 2, 3, 3, 1, 1} and the node number sequence {101, 102, 103, 104, 105, 106, 107, 108}, the path constructed by the system is a two-dimensional path structure combining the direction sequence and node numbers. The historical path samples formed by all abnormal batches confirmed through quality re-inspection are summarized in Table 4. Table 4 Sample Table of Historical Abnormal Trajectory Paths

[0040] As shown in Table 4, the system can establish a dual trajectory structure path of direction and node based on the image offset order and the trajectory structure of process nodes, and finally generate a set of historical abnormal trajectory paths.

[0041] The trend path matching submodule compares the similarity between the trend segment direction change and the historical trajectory direction evolution order based on the historical abnormal trajectory path set and the image offset trend segment sequence. It also performs position matching based on the time window of the trend segment and the process node number in the historical trajectory, filters the trajectories with consistent direction patterns and node positions, and generates a trend path matching result set. By combining historical abnormal trajectory path sets and image offset trend segment sequences, the system first extracts the dominant direction sequence and start-end time window within the frame sequence range for each segment in the trend segment sequence. Then, it compares this with the direction evolution sequence in the historical path set, performing direction order matching through window alignment. The matching process uses a direction matching scoring mechanism to calculate the matching degree. The scoring rule is: if the number of consecutive directions in the trend segment direction sequence that match the starting segment of the sequence in the historical path is greater than 3, then the direction sequence is considered to match. Furthermore, the system maps the image time interval to which the trend segment belongs to the node number sequence in the historical path to determine whether the node where the trend segment image is located is continuously covered in the historical path node sequence. Historical paths that meet both conditions are identified as potential evolutionary path matches. For example, if the trend segment frame sequence range is IMG0201 to IMG0210, the direction sequence is {1, 2, 2, 3, 3, 1}, and the corresponding node numbers are {105, 106, 107, 108, 109, 110}, and the historical path contains the direction sequence {1, 2, 2, 3} and the node number sequence {104, 105, 106, 107, 108}, then the path is identified as a matchable path. The system records the matching path number, matching score, trend segment number, and other information, sorts them by matching score, outputs all matches, and finally generates a trend path matching result set.

[0042] The evolution channel filtering submodule performs trajectory continuity judgment on all matching items based on the trend path matching result set, filters paths that meet the requirements of the same direction change sequence and continuous node number distribution, outputs the matching path number, associated frame sequence and process node sequence in a structured manner, and generates a quality evolution channel path list. Based on the trend path matching result set, the system extracts all matching path numbers and performs continuity verification to determine whether the direction sequence involved in the matching path is completely consistent with the direction sequence in the trend segment. It further verifies whether the node number distribution is continuous and whether the difference in number between adjacent nodes does not exceed 1. Paths that do not meet the above two conditions are eliminated, and only those that meet the conditions are retained. Simultaneously, each retained path is re-encoded to generate an evolution channel number, such as EP001, EP002, etc., and a structured path record table is constructed. The table contains fields such as channel number, direction mode, associated frame sequence, time interval, and node distribution, as shown in the example below: Table 5 Quality Evolution Channel Path Output Table

[0043] As shown in Table 5, the output record structure is clear and traceable. The content of each field comes from the item-by-item verification and coding matching of the trend path and historical path. Finally, the system generates a list of quality evolution channel paths.

[0044] Specifically, such as Figure 2 , 7 As shown, the collaborative traceability and aggregation module includes: The node path localization submodule is based on the quality evolution channel path list. It reads the node number sequence and the batch number to which the channel belongs in each channel path, establishes a node sequence index structure according to the path number and time order, performs node position parsing operation on each path and identifies the channel mapping relationship, and generates a channel node position index table. Based on the quality evolution channel path list, the system sequentially reads the process node number and batch number recorded in each path, calls the process flow database to extract the detection station, detection process name, and node position sequence corresponding to the node number, and establishes an index mapping table for the path number and node sequence. For example, if path EP003 contains node numbers {103, 104, 105, 106}, the system constructs the structure EP003→{103, 104, 105, 106}. The system further analyzes the position segment and function type of each node in the entire production process flow, and verifies whether the nodes belong to the same batch of production tasks in conjunction with the production plan table. For cross-batch cases, a separation marking process is performed. The system binds the node index structure of each path to the batch number and outputs the results to facilitate subsequent chain construction and invocation. Each node number in the node index structure corresponds to the equipment number and the sequence number of the process segment. The position index structure is generated according to time sequence, as shown in the example below. Table 6 Channel Node Location Index

[0045] As shown in Table 6, the system has completed the structural binding of channel paths with node numbers, equipment numbers, and process sections, and finally generated a channel node location index table.

[0046] The batch chain integration submodule extracts the production line to which the equipment belongs based on the channel node location index table and the corresponding equipment number mapped by the node number. It then connects the equipment that the path passes through in the same batch in chronological order and establishes corresponding structured chain information for different batches, generating a quality traceability chain structure set. After obtaining the channel node location index table, the system searches the equipment deployment database for the production line, equipment model, and deployment location information of each path node based on the equipment number field corresponding to each node in the table. For example, equipment D-03 to D-06 all belong to production line A-01 and are deployed in workstation areas Z3 to Z6. The system aggregates and identifies the equipment according to its production line. Equipment involved in each channel within the same batch is grouped and categorized uniformly. The system performs linear concatenation processing based on the order of equipment positions in the production line and the order of node numbers in the path to generate a quality traceability chain node list for this batch. At the same time, it adds fields for each node's detection time window, image number, and offset direction information to complete the data structure encapsulation. For batch B0071, the system constructs the following chain record: Table 7 Quality Traceability Chain Structure Record Table

[0047] As shown in Table 7, the system has completed the construction of quality traceability chain records by batch dimension, and finally generated a quality traceability chain structure set.

[0048] The end-to-cloud synchronous display submodule encapsulates the quality traceability chain structure set into a standard JSON format data structure required by the display interface, uploads it to the cloud display platform by batch dimension, generates a traceability visualization display page through data interface calls, and synchronously writes it to the local archive database to generate automotive fastener quality data traceability records. Based on the quality traceability chain structure set, the system transforms the chain structure of each batch into a standardized JSON data structure. The fields include batch number, chain node array, image number corresponding to each node, offset direction, detection equipment number, and process segment information. The "batch number" field represents the production batch number to which the current chain belongs. The "chain node array" field contains multiple subfields; each node includes five items: "node number," "equipment number," "process segment number," "offset direction code," and "image number." The system encapsulates all node content into a chain relationship in a structured list format. In the example, if the batch number is B0071, and the corresponding chain node content is nodes 103 to 106, then the encapsulated fields are as follows: Batch number B0071... 071. The chain node array sequentially includes: node number 103, equipment number D-03, process segment number P1, offset direction left, image number IMG2031; node number 104, equipment number D-04, process segment number P2, offset direction left, image number IMG2032; node number 105, equipment number D-05, process segment number P3, offset direction unbiased, image number IMG2033; and node number 106, equipment number D-06, process segment number P4, offset direction right, image number IMG2034. After the system completes field encapsulation, it organizes all traceability chain data by batch dimension and uploads it to the cloud visualization platform interface, which uses standard HTTP. The system uses a POST method for interaction. The uploaded content includes a chain structure and a timestamp field for the corresponding batch. After successful reception by the cloud, a status code of 200 and a link to the display page are returned. The system simultaneously writes this link to the local database table, with the fields consisting of batch number, visualization link, and upload timestamp. The upload status is recorded and archived, ultimately generating a traceability record for automotive fastener quality data.

[0049] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A cloud-edge collaborative automotive fastener quality data traceability system, characterized in that, include: The edge offset analysis module acquires images of automotive fasteners, reads the pixel coordinates of the bolt outline boundary and the template boundary in the image, calculates the pixel difference between the horizontal and vertical positions, filters the set of continuous boundary points, marks the abnormal offset state images, and outputs the offset state image set. Based on the offset state image set, the traceability tag binding module extracts the timestamp, device number and batch of the abnormal offset state image, combines and encodes them and binds them to the image tag information, and pushes them to the cloud tagging interface to generate a time batch binding tag set. The sequence trend construction module performs clustering identification and sequential pattern organization on the continuous distribution of image offset direction according to the time batch and the tag set, filters out segments with obvious offset trend changes and performs clustering and classification processing to generate image offset trend segment sequence. The channel path recognition module extracts historical inspection records of automotive fasteners, analyzes and determines historical abnormal evolution channel paths, and matches and identifies the current segment's directional distribution by combining the image offset trend segment sequence with the current segment's directional distribution. It then filters evolution channels that match the offset change direction and have consistent trajectory distribution positions, generating a quality evolution channel path list. The collaborative traceability and aggregation module locates the process node number corresponding to each path according to the quality evolution channel path list, integrates them into a quality data traceability chain by batch according to the equipment distribution, and displays them synchronously on the end-to-cloud interface to obtain automotive fastener quality data traceability records.

2. The end-cloud coordinated automotive fastener quality data traceability system of claim 1, wherein: Specifically, the abnormal offset state image is determined by comparing the differences in the continuous boundary point set with the set boundary offset recognition threshold, and judging whether there are more than 5 consecutive boundary points in each direction whose offsets in the same direction exceed the set boundary offset recognition threshold. If the corresponding images in both the horizontal and vertical directions are satisfied at the same time.

3. The end-to-cloud collaborative automotive fastener quality data traceability system according to claim 1, characterized in that: The segment with a significant change in offset trend is specifically defined as the segment in which the frequency of change in offset direction between adjacent frames is extracted, and the number of times the direction changes occur within every 10 frames and the span of the direction change is combined to identify segments in consecutive frames that have more than 6 instances of direction change and span changes of more than two direction categories.

4. The end-cloud coordinated automotive fastener quality data traceability system of claim 1, wherein: The specific path of the historical anomaly evolution channel is as follows: extract the image sequence marked in the batches that have been confirmed by quality re-inspection in the historical detection records, and construct the corresponding trajectory path according to the offset direction order of the images on the time axis and the process node number to which they belong.

5. The end-cloud coordinated automotive fastener quality data traceability system of claim 1, wherein: The offset state image set includes offset direction classification markers, abnormal state image indexes, and pixel difference distribution maps. The time batch binding tag set includes image time indexes, equipment station numbers, and abnormal batch codes. The image offset trend segment sequence includes offset direction change trajectories, change frequency statistical features, and trend pattern grouping categories. The quality evolution channel path list includes direction sequence paths, process node mappings, and image sequence index mappings. The quality data traceability record includes manufacturing node marker chains, batch traceability mapping chains, and end-to-cloud display interface links.

6. The end-cloud coordinated automotive fastener quality data traceability system of claim 1, wherein, The edge offset parsing module includes: The image boundary extraction submodule acquires images of automotive fasteners, extracts the pixel coordinates of the bolt contour boundary and the template boundary in the image, performs contour recognition on the bolt image, obtains the actual boundary point set in the image through coordinate mapping, and generates an image boundary coordinate set. The position deviation calculation submodule reads the horizontal and vertical pixel coordinate axes to construct a boundary comparison index set based on the image boundary coordinate set, calculates the pixel difference between the bolt outline boundary and the template boundary in the horizontal and vertical directions, and performs a filtering operation on the continuous boundary point region in the difference sequence to obtain a continuous direction deviation difference sequence. The offset anomaly identification submodule performs a continuity threshold judgment operation on the horizontal and vertical directions based on the continuous directional deviation difference sequence. It compares each value in the difference sequence with the set boundary offset identification threshold, marks the abnormal offset state image, and generates an offset state image set.

7. The end-cloud coordinated automotive fastener quality data traceability system of claim 1, wherein, The traceability tag binding module includes: The image information extraction submodule, based on the offset state image set, reads the timestamp and device number by embedding metadata in the image file header, extracts the batch number information by matching the production batch record through the image path, and generates an image traceability information group; The label content generation submodule, based on the image traceability information group, concatenates the timestamp, device number, and batch number to construct a traceability label string in a unified format, writes it into the corresponding label of the image, and generates a time-batch combined label dataset. The cloud-based binding and push submodule encapsulates the unique image identifier and combined label into a standard data structure based on the time-batch combined label dataset, constructs a data upload request body, pushes data through the cloud tagging interface and receives binding response information, and generates a time-batch binding label set.

8. The end-cloud coordinated automotive fastener quality data traceability system of claim 1, wherein, The sequence trend construction module includes: The orientation distribution recognition submodule extracts the offset orientation identifier corresponding to each frame of the image according to the time batch binding tag set, forms a continuous sequence of offset orientation information of adjacent image frames, arranges them in chronological order to form an orientation distribution chain, groups and counts the repetitive orientation segments, and generates an image orientation distribution coding sequence. The frequency feature extraction submodule, based on the image orientation distribution encoding sequence, counts the number of orientation changes by sliding window, calculates the orientation number span value within each window, records the segments with significant changes, outputs the start and end frame number indices of the segments, and generates a set of orientation change frequency indicators. The trend segment clustering submodule extracts image frame segments that continuously meet the conditions of direction change frequency and span based on the set of direction change frequency indicators. It performs temporal adjacency and direction encoding feature similarity matching on each segment, groups the related segments into the same group, and outputs the corresponding set of image frame numbers to generate an image offset trend segment sequence.

9. The end-cloud coordinated automotive fastener quality data traceability system of claim 1, wherein, The channel path identification module includes: The abnormal trajectory extraction submodule obtains the image sequence of abnormal batches confirmed by quality re-inspection from the historical inspection records, extracts the timestamp, offset direction and corresponding process node number of each image, sorts them in ascending order by image timestamp, outputs the direction evolution order and the corresponding structure of process node index, and generates a set of historical abnormal trajectory paths. The trend path matching submodule compares the similarity between the trend segment direction change and the historical trajectory direction evolution order based on the historical abnormal trajectory path set and the image offset trend segment sequence. It also performs position matching based on the time window of the trend segment and the process node number in the historical trajectory, filters trajectories with consistent direction patterns and node positions, and generates a trend path matching result set. The evolution channel filtering submodule performs trajectory continuity judgment on all matching items based on the trend path matching result set, filters paths that meet the requirements of the same direction change sequence and continuous node number distribution, outputs the matching path number, associated frame sequence and process node sequence in a structured manner, and generates a quality evolution channel path list.

10. The end-cloud coordinated automotive fastener quality data traceability system of claim 1, wherein, The collaborative traceability and aggregation module includes: The node path localization submodule, based on the quality evolution channel path list, reads the node number sequence and the batch number to which the channel belongs in each channel path, establishes a node sequence index structure according to the path number and time order, performs node position parsing operation on each path and identifies the channel mapping relationship, and generates a channel node position index table. The batch chain integration submodule extracts the production line to which the equipment belongs based on the channel node position index table and the corresponding equipment number mapped by the node number. It then connects the equipment that the path passes through in the same batch in chronological order and establishes corresponding structured chain information for different batches to generate a quality traceability chain structure set. The end-to-cloud synchronous display submodule encapsulates the quality traceability chain structure set into a standard JSON format data structure required by the display interface, uploads it to the cloud display platform by batch, generates a traceability visualization display page through data interface calls, and synchronously writes it to the local archive database to generate automotive fastener quality data traceability records.

Images