A tensioner detection system and method based on line-product coupling

Abstract

This invention discloses a tensioner detection system and method based on production line-product coupling, belonging to the field of tensioner detection technology. Based on spatial overlap analysis results and temporal entry information, the system determines the attribution of feature elements within the overlap area. A comprehensive matching score is calculated based on the spatial proximity and temporal matching degree of the feature elements, and the feature elements are assigned to the inspection target with the highest comprehensive matching score. During the inspection of the inspection target, the system automatically scans and analyzes all potential workstation gaps on the current production line that can be used for inspection tasks. It analyzes whether candidate positions have temporal conflicts with existing inspection tasks. If a conflict exists, the avoidance coefficient of each candidate position is obtained, and the candidate position with the smallest avoidance coefficient is selected as the inspection intervention point to inspect the inspection target. This detection system achieves accurate detection of tensioners in complex production line environments, significantly improving the intelligence level and overall efficiency of the inspection process.

Description

Technical Field

[0001] This invention relates to the field of tensioner testing technology, specifically to a tensioner testing system and method based on production line-product coupling. Background Technology

[0002] In the automated inspection processes of modern industrial production lines, efficient and accurate inspection of key components such as tensioners faces numerous challenges. Traditional inspection methods often struggle to effectively handle inspection scenarios where multiple inspected targets are operating concurrently on the production line, and where the spatial distribution of feature elements is complex and dynamically changing over time. Especially when the inspection fields of multiple inspected targets overlap, problems such as misclassification of feature element attribution and low inspection efficiency easily arise, affecting overall inspection accuracy and production line smoothness. Therefore, there is an urgent need for a tensioner inspection method that can combine the coupling characteristics of the production line and product, and achieve accurate clustering of feature elements and intelligent selection of inspection intervention points through collaborative analysis of spatial distribution and temporal information. Summary of the Invention

[0003] The purpose of this invention is to provide a tensioner detection system and method based on production line-product coupling, so as to solve the problems in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a tensioner detection method based on production line-product coupling, the detection method comprising the following steps:

[0005] A1: Collect the content to be inspected, which covers multiple inspected targets, each of which consists of several feature elements;

[0006] A2: Based on the spatial distribution characteristics of the feature elements in each inspected target and combined with the temporal entry information, preliminary clustering is performed on all feature elements, and feature elements belonging to the same inspected target are aggregated into a feature set to form a preliminary detection feature profile of the inspected target.

[0007] A3: When the detection field of view of any two inspected targets overlaps, based on the spatial overlap analysis results and temporal entry information, the attribution judgment is carried out on the feature elements within the overlapping area. The comprehensive matching score is calculated based on the spatial proximity and temporal matching degree of the feature elements, and the feature elements are assigned to the inspected target with the highest comprehensive matching score.

[0008] A4: When performing inspection operations on the target under inspection, automatically scan and analyze all potential workstation gaps on the current production line that can be used for inspection tasks as candidate positions. Analyze whether the candidate positions have timing conflicts with existing inspection tasks. If there is a conflict, obtain the avoidance coefficient of each candidate position and select the candidate position with the smallest avoidance coefficient as the inspection intervention point to inspect the target under inspection.

[0009] Furthermore, the attribution of feature elements within the overlapping area is determined, including the entry sequence of the two inspected targets and the relative positional relationship between the feature within the overlapping area and the visual boundaries of the two inspected targets.

[0010] Furthermore, the feature elements are assigned to the tested targets with the highest overall matching scores, including the following steps:

[0011] Based on the predefined detection field of view boundary, the overlapping area of ​​the fields of view of any two inspected targets is calculated, and the overlapping area is determined by geometric operations in coordinate space;

[0012] Filter all inspected targets whose feature coordinates are located within the overlapping area to form a set of feature elements to be judged, S_overlap;

[0013] Iterate through each feature element s_i in the feature set S_overlap, calculate its comprehensive matching score with all relevant inspected targets, and assign s_i to the inspected target with the highest comprehensive matching score.

[0014] Furthermore, the spatial proximity is the Euclidean distance between the feature element and the center of the detection field of view of the inspected target; the smaller the distance, the stronger the spatial correlation.

[0015] The temporal matching degree is the degree to which the timing of the appearance of a feature element matches the timing of the entry of a certain inspected target. The higher the matching degree, the closer the temporal correlation.

[0016] Furthermore, the comprehensive matching score is calculated based on the spatial proximity and temporal matching degree of the feature elements, including the following steps: the spatial proximity and temporal matching degree are respectively subjected to maximum-minimum normalization processing, and the normalized spatial proximity and temporal matching degree are weighted to calculate the comprehensive matching score between each feature element and each test subject, where spatial proximity is an inverse proportional factor and temporal matching degree is a direct proportional factor;

[0017] Spatial proximity index: For example, the Euclidean distance between the feature element and the center of the detection field of view of the detected target (such as the geometric center of the ROI). The smaller the distance, the stronger the spatial correlation.

[0018] Temporal matching degree index: For example, the degree to which the timing of the occurrence of this feature element matches the timing of the entry of a certain inspected target. The higher the matching degree, the closer the temporal correlation.

[0019] Furthermore, the spatial proximity is used to quantify the spatial correlation strength between the feature element and the detection field center of a certain inspected target. For each inspected target, its center coordinates are calculated through the geometric centers of all clustered feature elements of the inspected target. For each feature element s_i in the feature element set S_overlap, the Euclidean distance d_ij between it and the field center C_j of the inspected target T_j is calculated as the spatial proximity.

[0020] Furthermore, the time-series matching degree is used to quantify the temporal correlation between the occurrence time of a feature element and the entry time of a certain inspected target. The entry time t_Tj of each inspected target is determined by its actual arrival timestamp on the production line. The time sequence t_si of feature element s_i is associated with its original detection data stream. For each feature element s_i in the feature element set S_overlap, the absolute deviation Δt_ij between its time sequence t_si and the entry time t_Tj of the inspected target T_j is calculated, and then the absolute deviation Δt_ij is converted into the time-series matching degree.

[0021] Furthermore, the avoidance coefficient for each candidate position is obtained. The calculation logic for the avoidance coefficient is as follows:

[0022] Based on the detection criticality index of the detection task in the production line and the overall cycle time impact index, an avoidance priority index is generated. The larger the avoidance priority index, the more the candidate position needs to avoid the current detection task. The avoidance coefficient is obtained by summing the avoidance priority indices of all detection tasks at each candidate position.

[0023] Furthermore, when performing inspection operations on the target under inspection, the system automatically scans and analyzes all potential workstation gaps on the current production line that are available for inspection tasks, using them as candidate positions. It then analyzes whether these candidate positions conflict with existing inspection tasks in terms of timing. If a conflict exists, it obtains the avoidance coefficient for each candidate position and selects the candidate position with the smallest avoidance coefficient as the inspection intervention point to inspect the target under inspection. This includes the following steps:

[0024] If the time range of a candidate location overlaps with the execution time range of a scheduled detection task, it is determined that the candidate location and the corresponding detection task have a timing conflict.

[0025] For each detection task that has a timing conflict with the candidate position, the avoidance priority index of the detection task is calculated based on its criticality index in the production line and the overall cycle time impact index.

[0026] The detection criticality index and the overall beat impact index are normalized by the maximum and minimum values. The avoidance priority index is obtained by summing the normalized detection criticality index and the overall beat impact index.

[0027] For each candidate location, the avoidance priority index of all conflicting detection tasks is aggregated, and the avoidance coefficient of the candidate location is calculated by summing the avoidance priority indices of all detection tasks.

[0028] Specifically, this application also provides a tensioner detection system based on production line-product coupling, including an overlap area analysis module, an attribution judgment module, and an avoidance analysis module;

[0029] Overlapping Area Analysis Module: Collects the content to be inspected, which covers multiple inspected targets. Each inspected target is composed of several feature elements. Based on the spatial distribution characteristics of the feature elements in each inspected target and combined with the temporal entry information, preliminary clustering is performed on all feature elements. Feature elements belonging to the same inspected target are aggregated into a feature set to form a preliminary detection feature profile of the inspected target. When the detection field of view of any two inspected targets overlaps, a discrimination criterion is introduced to analyze the inspected target affiliation relationship of the feature elements in the overlapping area.

[0030] Attribution determination module: For overlapping areas, based on spatial overlap analysis results and temporal entry information, attribution determination is carried out on feature elements within the overlapping area. The comprehensive matching score is calculated based on the spatial proximity and temporal matching degree of the feature elements, and the feature elements are assigned to the inspected target with the highest comprehensive matching score.

[0031] Avoidance Analysis Module: When performing inspection operations on the target under inspection, it automatically scans and analyzes all potential workstation gaps on the current production line that can be used for inspection tasks as candidate positions. It analyzes whether the candidate positions have timing conflicts with existing inspection tasks. If there is a conflict, it obtains the avoidance coefficient of each candidate position and selects the candidate position with the smallest avoidance coefficient as the inspection intervention point to inspect the target under inspection.

[0032] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0033] This application automatically scans potential workstation gaps on the production line as candidate locations and intelligently analyzes their timing conflicts with existing inspection tasks. It selects the optimal inspection intervention point with the lowest avoidance coefficient, effectively preventing mutual interference between inspection tasks and optimizing the inspection timing, thereby improving the overall operational efficiency of the production line and the smoothness of the inspection process. In summary, this solution, through deep coupling between the production line and the product, combined with multi-dimensional spatial and temporal analysis, achieves precise clustering of feature elements, accurate attribution of overlapping areas, and intelligent selection of inspection intervention points during the tensioner inspection process, significantly improving inspection accuracy, efficiency, and production line collaboration capabilities.

[0034] This application utilizes the spatial distribution characteristics and temporal entry information of feature elements to perform preliminary clustering of feature elements, aggregating feature elements of the same inspected target into a feature set to form a preliminary detection feature profile. This not only effectively reduces data redundancy but also preliminarily constructs the feature contour of each inspected target, laying the foundation for accurate detection. At the same time, for feature elements in overlapping areas of the detection field of view, a special discrimination criterion is introduced to ensure accurate analysis of the attribution relationship of the inspected targets, avoiding misjudgment and missed detection.

[0035] This application achieves accurate attribution of feature elements within overlapping regions by deeply analyzing feature elements in the overlapping regions, calculating a comprehensive matching score based on spatial proximity and temporal matching degree, and assigning feature elements to the highest-scoring target, thereby significantly improving the accuracy and reliability of detection. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0037] Figure 1 This is a flowchart of the detection method of the present invention.

[0038] Figure 2 This is a framework diagram of the detection system of the present invention. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] Example: This example provides a tensioner detection method based on production line-product coupling. Please refer to [link / reference]. Figure 1 As shown, the detection method includes the following steps:

[0041] A1: At a designated inspection station on an industrial production line, the system first collects and presents the content to be inspected that needs to be processed. This content covers multiple inspected targets (such as multiple tensioner assemblies), and each inspected target consists of several key feature elements.

[0042] In the actual spatial layout, these inspected targets may be located adjacently or have overlapping detection fields (e.g., multiple tensioners integrated into the same component, or products densely arranged on the conveyor line). Simultaneously, the timing of each inspected target's arrival is recorded. This timing information typically depends on the product's arrival order on the production line, serving as a time reference for subsequent logical reasoning and grouping.

[0043] A2: Based on the spatial distribution characteristics of key feature elements in each inspected target (such as relative spacing, azimuth angle, arrangement pattern, etc.) and combined with their temporal entry information, preliminary clustering is performed on all feature elements. The aim is to aggregate feature elements belonging to the same inspected target (such as the same tensioner) into a unified feature set, forming a preliminary detection feature profile of the inspected target.

[0044] A3: In particular, when the detection fields of two inspected targets (such as tensioner A and tensioner B) overlap (e.g., the region of interest (ROI), detection window, or spatial coverage in the image), the system does not rely solely on spatial arrangement information, but further introduces the following criteria:

[0045] The relative positional relationship between the arrival sequence of the two inspected targets (i.e., which inspected target arrives at the inspection station first) and the key features (such as specific markers, edge contours, structural markings, etc.) located in the overlapping area and the visual field boundary of the two inspected targets;

[0046] The relationship between the inspected target and the feature elements in the overlapping area is analyzed to effectively resolve the common problems of feature element mismatch or ambiguous attribution in multi-target detection scenarios, and to ensure that each feature element is correctly bound to its inspected target.

[0047] For overlapping areas (such as two tensioners partially superimposed or overlapping detection fields), based on the spatial overlap analysis results and temporal entry information, the attribution of feature elements (such as mounting holes, markings, thread structures, etc.) within the overlapping area is determined.

[0048] This judgment process incorporates the following two key indicators and performs standardized and weighted fusion calculations:

[0049] Spatial proximity index: For example, the Euclidean distance between the feature element and the center of the detection field of view of the detected target (such as the geometric center of the ROI). The smaller the distance, the stronger the spatial correlation.

[0050] Temporal matching degree index: For example, the degree to which the timing of the occurrence of this feature element matches the timing of the entry of a certain inspected target. The higher the matching degree, the closer the temporal correlation.

[0051] The system performs max-min normalization on these two indicators and sets reasonable weight configurations (usually spatial proximity is the inverse factor and temporal matching degree is the direct factor), and then calculates the comprehensive matching score between each feature element and each inspected object; finally, the feature element is assigned to the inspected object with the highest comprehensive matching score, thereby achieving accurate binding of feature elements in the overlapping area.

[0052] A4: When performing inspection operations on tensioners, automatically scan and analyze all potential workstation gaps on the current production line that are available for inspection tasks (e.g., idle time periods after a certain process) as possible candidate locations. For each potential workstation gap, evaluate the following factors in sequence:

[0053] Does the candidate location conflict with existing detection tasks in terms of timing (e.g., overlapping detection actions)? If a conflict exists, the avoidance priority of the affected detection tasks is further identified (based on the detection criticality index of the detection task in the production line and the overall cycle time impact index), and an avoidance priority index is generated. The larger the avoidance priority index, the more the candidate location needs to avoid the current detection task. The avoidance priority indices of all detection tasks at each candidate location are summed to obtain the avoidance coefficient. The candidate location with the smallest avoidance coefficient is selected as the detection intervention point to achieve accurate detection of the tensioner status, while ensuring that the main production process and other key detection tasks are not affected or have minimal impact.

[0054] This embodiment also provides a tensioner detection system based on production line-product coupling. Please refer to [link / reference]. Figure 2 As shown, it includes an overlapping area analysis module, an attribution determination module, and an avoidance analysis module;

[0055] Overlapping Area Analysis Module: Collects the content to be inspected, which covers multiple inspected targets. Each inspected target is composed of several feature elements. Based on the spatial distribution characteristics of the feature elements in each inspected target and combined with the temporal entry information, preliminary clustering is performed on all feature elements. Feature elements belonging to the same inspected target are aggregated into a feature set to form a preliminary detection feature profile of the inspected target. When there is an overlapping area in the detection field of any two inspected targets, a discrimination criterion is introduced to analyze the inspected target attribution relationship of the feature elements in the overlapping area. The overlapping area analysis results are sent to the attribution judgment module.

[0056] Attribution Module: For overlapping areas, based on spatial overlap analysis results and temporal entry information, attribution judgment is performed on feature elements within the overlapping area. A comprehensive matching score is calculated based on the spatial proximity and temporal matching degree of the feature elements, and the feature elements are assigned to the inspected target with the highest comprehensive matching score. The inspected target after attribution is sent to the avoidance analysis module.

[0057] Avoidance Analysis Module: When performing inspection operations on the target under inspection, it automatically scans and analyzes all potential workstation gaps on the current production line that can be used for inspection tasks as candidate positions. It analyzes whether the candidate positions have timing conflicts with existing inspection tasks. If there is a conflict, it obtains the avoidance coefficient of each candidate position and selects the candidate position with the smallest avoidance coefficient as the inspection intervention point to inspect the target under inspection.

[0058] The following is a detailed description of each step in this application:

[0059] A1: At a designated inspection station on an industrial production line, the current inspection content to be processed is first collected and presented. This content covers multiple inspected targets (such as multiple tensioner assemblies), and each inspected target consists of several key feature elements.

[0060] In the actual spatial layout, these inspected targets may be located adjacently or have overlapping detection fields (e.g., multiple tensioners integrated into the same component, or products densely arranged on the conveyor line). Simultaneously, the timing of each inspected target's arrival is recorded. This timing information typically depends on the product's arrival order on the production line, serving as a time reference for subsequent logical reasoning and grouping.

[0061] Because products on the production line may be densely arranged or integrated (e.g., multiple tensioner assemblies fixed to the same metal bracket, or products closely arranged on a conveyor belt), these inspected targets may be adjacent in position or have partially overlapping fields of view in the actual spatial layout. For example, when two tensioner assemblies are adjacent on the conveyor belt with a small gap, their corresponding fields of view may partially overlap due to overlapping camera views. This can cause some feature elements (such as screws and buckles) to be captured by multiple cameras simultaneously, leading to ambiguity in the identification of the inspected target.

[0062] The system synchronously records the timing of each inspected target's entry. This information is typically based on the product's arrival order on the production line (such as the sequence on the conveyor belt or the timing of the robotic arm's grasping), and serves as the time reference for subsequent logical reasoning and grouping. For example, if two tensioner components pass through the same inspection station sequentially, their order can be determined based on their entry timestamps (such as millisecond-level timestamps), thereby assisting in subsequent feature clustering and attribution analysis.

[0063] A2: Based on the spatial distribution characteristics of key feature elements in each inspected target (such as relative spacing, azimuth angle, arrangement pattern, etc.) and combined with their temporal entry information, preliminary clustering is performed on all feature elements. The aim is to aggregate feature elements belonging to the same inspected target (such as the same tensioner) into a unified feature set, forming a preliminary detection feature profile of the inspected target.

[0064] After collecting the data to be inspected, the preliminary clustering stage begins. The inspected target is aggregated by fusing the spatial distribution characteristics and temporal entry information of key feature elements. Feature elements belonging to the same inspected target (such as a single tensioner component) are grouped into a unified feature set, constructing a preliminary detection feature profile for that target and providing structured input for subsequent accurate detection. Simultaneously, to address the unavoidable problem of overlapping detection fields of view in multi-target detection (such as multiple tensioners intersecting in the imaging field due to dense arrangement or integrated assembly), multi-dimensional discrimination criteria are introduced to analyze the target attribution relationship of feature elements within overlapping areas, avoiding detection errors caused by feature mismatch or ambiguous attribution.

[0065] The collected content to be inspected is deconstructed to identify that each inspected target consists of several key feature elements. These feature elements may be geometric structural features (such as the bolt hole position of the tensioner, the coordinate position of the buckle boss), topological connection relationship (such as the relative spatial layout of adjacent parts), surface texture features (such as the reflectivity distribution of a specific area), or assembly markings (such as the position of the engraved product serial number).

[0066] Each feature is labeled with its absolute coordinates in three-dimensional space (or two-dimensional imaging plane) (such as (x,y,z) in the camera coordinate system or planar coordinates in the conveyor line coordinate system), and the timing entry information of the target to which it belongs is recorded synchronously (for example, based on the pulse count of the conveyor belt encoder or the timing record of the robotic arm grasping, the order in which each tensioner arrives at the inspection station is determined, usually represented by a timestamp t_i).

[0067] A clustering algorithm driven by both spatial distribution characteristics and temporal information is used to initially group all feature elements. The specific logic is as follows:

[0068] For each feature element, calculate its relative spatial relationship with other feature elements, including but not limited to:

[0069] Relative distance (e.g., Euclidean distance d_ab = √[(x_a-x_b)]) 2 + (y_a-y_b) 2 + (z_a-z_b) 2 ], where (x_a, y_a, z_a) represents the three-dimensional coordinates of feature element a, and (x_b, y_b, z_b) represents the three-dimensional coordinates of feature element b, used to determine whether feature elements are in a close adjacency state; azimuth angle (such as the angle θ between the line connecting two feature elements and the reference coordinate axis, used to analyze the relative orientation of the parts); arrangement pattern (such as whether multiple feature points are linear, circular or symmetrically distributed).

[0070] Combining the timing entry information (t_i, t_j), features that arrive simultaneously or in consecutive time (such as multiple tensioner assemblies belonging to the same batch of products) are preferentially included in the same cluster candidate set. For example, if the timing entry time difference between two bolt hole positions is less than the threshold Δt (e.g., 100ms) and their spatial distance is less than the preset spacing δ for similar components (e.g., 5mm), then they are initially considered to belong to the same inspected target.

[0071] Through the above spatial-temporal joint analysis, feature elements that meet the following conditions are aggregated into the same feature set:

[0072] (1) The spatial relative spacing is less than the spacing δ, and the azimuth and arrangement pattern conform to the structural characteristics of the inspected target (e.g., the bolt holes of the tensioner are usually distributed in a triangle with a spacing of about 3-5 mm); (2) They belong to the same arrival batch in time (time difference ≤ Δt), or the temporal continuity with the clustered feature elements is logical (e.g., the spatial relationship between the subsequently arriving component features and the clustered set is reasonable). Finally, each feature set corresponds to a preliminary detection feature profile of an inspected target, which includes the coordinates, topological relationships, and temporal labels of all feature elements of the inspected target.

[0073] When two or more inspected targets (such as tensioner A and tensioner B) have overlapping fields of view due to dense spatial layout or integrated assembly (such as multiple tensioners fixed on the same metal bracket), some feature elements (such as buckle protrusions with shared boundaries, adjacent bolt holes) may fall within the field of view of multiple inspected targets at the same time. In this case, relying solely on spatial distribution characteristics (such as relative spacing) may lead to misclassification (for example, misclassifying the buckle of tensioner B as a component of tensioner A).

[0074] To address this issue, multi-dimensional criteria are introduced during the overlapping region analysis phase, specifically including:

[0075] Consider the arrival order of the inspected targets to which the feature elements belong. For example, if the timing of tensioner A's arrival time t_A is earlier than that of tensioner B's arrival time t_B (i.e., A arrives at the inspection station first), then the feature elements (such as the fixing screws on the bracket) that are closer to the start of the conveyor line (or upstream station) in the overlapping area are more likely to belong to the tensioner A that arrived first.

[0076] Analyze the geometric relationship between key features (such as specific markers, edge contours, and structural markings) within the overlapping area and the visual boundaries of the two inspected targets. For example, if a marker within the overlapping area (such as the unique engraved serial number of tensioner A) is located inside the visual boundary of tensioner A (i.e., closer to the center of A's visual field and more than the threshold d_threshold of B's ​​visual boundary), then the marker is determined to belong to tensioner A; if an edge contour (such as the outer edge of the buckle of tensioner B) has a higher degree of overlap with the visual boundary of tensioner B (e.g., the average distance between the contour point and the visual boundary of B is less than the average distance with the visual boundary of A), then it is likely to be assigned to tensioner B.

[0077] Further examine the logical relationship between the feature elements in the overlapping area and the core structure of the target under inspection. For example, if a bolt hole is spatially closer to a known core component (such as a ratchet mechanism) of tensioner A and farther from the core component of tensioner B, then the determination that it belongs to tensioner A is strengthened.

[0078] A3: In particular, when the detection field of view (such as the region of interest (ROI), detection window, or spatial coverage) of two inspected targets (e.g., tensioner A and tensioner B) overlaps, the following criteria are introduced instead of relying solely on spatial arrangement information:

[0079] The relative positional relationship between the arrival sequence of the two inspected targets (i.e., which inspected target arrives at the inspection station first) and the key features (such as specific markers, edge contours, structural markings, etc.) located in the overlapping area and the visual field boundary of the two inspected targets;

[0080] The analysis of the target attribution relationship of feature elements in overlapping areas can effectively resolve the common problems of feature element mismatch or ambiguity in multi-target detection scenarios, and ensure that each feature element is correctly bound to its respective target.

[0081] For overlapping areas (such as two tensioners partially superimposed or overlapping detection fields), based on the spatial overlap analysis results and temporal entry information, the attribution of feature elements (such as mounting holes, markings, thread structures, etc.) within the overlapping area is determined.

[0082] This judgment process incorporates the following two key indicators and performs standardized and weighted fusion calculations:

[0083] Spatial proximity index: For example, the Euclidean distance between the feature element and the center of the detection field of view of the detected target (such as the geometric center of the ROI). The smaller the distance, the stronger the spatial correlation.

[0084] Temporal matching degree index: For example, the degree to which the timing of the occurrence of this feature element matches the timing of the entry of a certain inspected target. The higher the matching degree, the closer the temporal correlation.

[0085] These two indicators are subjected to maximum-min normalization and a reasonable weight configuration is set (usually spatial proximity is the inverse factor and temporal matching degree is the direct factor). Then, the comprehensive matching score between each feature element and each subject is calculated. Finally, the feature element is assigned to the subject with the highest comprehensive matching score, thereby achieving accurate binding of feature elements in the overlapping area.

[0086] In multi-target inspection scenarios, when the inspection fields of two or more targets (e.g., two tensioners with partially overlapping structures due to integrated assembly or densely arranged conveyor lines) overlap (e.g., overlapping ROIs of industrial cameras, overlapping spatial coverage of multiple sensors, or overlapping field of view angles of 3D scanning equipment), some feature elements (such as mounting holes, markings, threaded structures, etc.) may simultaneously fall within the field of view of multiple targets. In this case, relying solely on a single spatial or temporal information may lead to misclassification of feature element attribution (e.g., misclassifying the mounting hole of target B as part of target A), thus affecting the accuracy of subsequent inspections. To address this issue, a standardized comprehensive matching score calculation logic is constructed by integrating two key indicators: spatial proximity and temporal matching, to achieve accurate binding of feature elements within overlapping areas.

[0087] Specifically, based on a predefined detection field of view boundary (such as the pixel-level ROI coordinate range of a single camera), the overlapping region (i.e., the overlapping area) of the fields of view of any two inspected targets (e.g., tensioner A and tensioner B) is calculated. This overlapping region is determined through geometric operations in coordinate space (e.g., the intersection polygon of two rectangles in a two-dimensional plane, or the common voxel set of multiple cuboids in three-dimensional space). Subsequently, all inspected targets whose feature coordinates lie within this overlapping region (e.g., the center point (x, y, z) of the mounting hole falls within the intersection region) are selected to form a set of feature elements to be judged, S_overlap. These feature elements include geometric structural features (e.g., the center coordinates of bolt holes), surface marking features (e.g., the pixel position of engraved marks), or topological connection features (e.g., the starting point coordinates of threaded structures).

[0088] Spatial proximity is used to quantify the spatial correlation strength between a feature element and the center of the detection field of view of a target subject. In other words, the smaller the distance between the feature element and the center of the target's field of view, the closer their spatial distribution, and the higher the probability that the feature element belongs to the target subject. The specific processing logic is as follows:

[0089] For each inspected target (such as tensioner A), its center coordinates are calculated using the geometric center of all clustered feature elements of that target (e.g., the mean point of all feature point coordinates, or the coordinate position of core structural components (such as ratchet mechanisms)). If the inspected target has not yet completed all feature clustering, the geometric center of its initial detection field of view (e.g., the center point of the camera ROI, or a preset assembly reference coordinate) is used.

[0090] For each feature element s_i (coordinates (x_i, y_i, z_i)) in the feature element set S_overlap, calculate the Euclidean distance d_ij (a physical quantity reflecting the degree of spatial proximity) between it and the center of the field of view C_j (coordinates (x_cj, y_cj, z_cj)) of the inspected target T_j (such as tensioner A).

[0091] Since the centers of vision of different inspected targets may be distributed in different spatial locations (for example, the center of tensioner A is closer to the front end of the conveyor line, and the center of tensioner B is closer to the rear end), directly using the original distance values ​​will lead to incomparable indicators. Therefore, d_ij is subjected to max-min normalization (mapping the original distance to the interval [0,1]), and the distance set {d_1j, d_2j, ..., d_nj} between all feature elements in the feature element set S_overlap and the inspected target T_j is traversed. The maximum value d_max_j and the minimum value d_min_j are calculated, and then d_ij is converted into a temporal matching degree SD_ij = (d_max_j - d_ij) / (d_max_j - d_min_j). At this time, the closer SD_ij is to 1, the smaller the spatial distance between feature element s_i and the inspected target T_j (the higher the closeness), and vice versa.

[0092] Temporal matching degree is used to quantify the temporal correlation between the occurrence time of a feature element and the arrival time of a target being inspected. In other words, the closer the temporal sequence of a feature element is to the actual arrival time of the target being inspected, the more consistent they are in temporal logic, and the higher the probability that the feature element belongs to the target being inspected. Specifically:

[0093] The arrival time t_Tj of each inspected target (such as tensioner A) is determined by its actual arrival timestamp on the production line (e.g., the counting time of the conveyor belt encoder, the timing record of the robotic arm grasping). The timing t_si of feature element s_i is associated with its original detection data stream (e.g., the timestamp of the image frame corresponding to the feature element, or the time of sensor acquisition).

[0094] For each feature element s_i in the feature element set S_overlap, calculate the absolute deviation Δt_ij = |t_si - t_Tj| between its temporal sequence t_si and the entry time t_Tj of the inspected target T_j (reflecting the time difference of temporal correlation). Similarly, perform max-min normalization on Δt_ij, iterating through the set of temporal deviations {Δt_1j, Δt_2j, ..., Δt_nj} between all feature elements in the feature element set S_overlap and the inspected target T_j, calculating the maximum value Δt_max_j and the minimum value Δt_min_j. Then, convert Δt_ij into a temporal matching degree TM_ij = 1 - (Δt_ij - Δt_min_j) / (Δt_max_j - Δt_min_j). Here, the closer TM_ij is to 1, the closer the temporal sequence of feature element s_i is to the entry time of the inspected target T_j (higher matching degree); conversely, the closer the temporal sequence is to 1, the lower the matching degree.

[0095] By weighted fusion of spatial proximity (SD_ij) and temporal matching degree (TM_ij), the comprehensive matching score MS_ij between each feature element s_i and each detected target T_j is calculated. Spatial proximity is an inverse factor (i.e., the smaller the distance, the higher the score; therefore, the standardized SD_ij is directly used as a positive contribution); temporal matching degree is a direct factor (i.e., the more matched the temporal sequence, the higher the score; therefore, the standardized TM_ij is directly used as a positive contribution). According to the needs of the actual detection scenario, weights are assigned to the two indicators (e.g., spatial proximity weight W_SD=0.6, temporal matching degree weight W_TM=0.4, reflecting the dominance of spatial correlation while taking into account the auxiliary judgment of temporal logic).

[0096] The calculation logic of the comprehensive matching score is described as follows: For each combination of s_i and T_j, the standardized SD_ij and TM_ij are summed according to their weights (for example, MS_ij = W_TM TM_ij-W_SD SD_ij), and finally a value between [0,1] is obtained. The higher the value, the stronger the association between s_i and T_j.

[0097] Iterate through each feature element s_i in the feature element set S_overlap, calculate its comprehensive matching score (MS_iA, MS_iB) with all relevant inspected targets (such as tensioner A and tensioner B), and assign s_i to the inspected target with the highest comprehensive matching score (for example, if MS_iA > MS_iB, then s_i is determined to belong to tensioner A).

[0098] A4: When performing inspection operations on tensioners, automatically scan and analyze all potential workstation gaps on the current production line that are available for inspection tasks (e.g., idle time periods after a certain process) as possible candidate locations. For each potential workstation gap, evaluate the following factors in sequence:

[0099] Does the candidate location conflict with existing detection tasks in terms of timing (e.g., overlapping detection actions)? If a conflict exists, the avoidance priority of the affected detection tasks is further identified (based on the detection criticality index of the detection task in the production line and the overall cycle time impact index), and an avoidance priority index is generated. The larger the avoidance priority index, the more the candidate location needs to avoid the current detection task. The avoidance priority indices of all detection tasks at each candidate location are summed to obtain the avoidance coefficient. The candidate location with the smallest avoidance coefficient is selected as the detection intervention point to achieve accurate detection of the tensioner status, while ensuring that the main production process and other key detection tasks are not affected or have minimal impact.

[0100] After completing the accurate clustering and attribution binding of the tensioner's feature elements, the inspection task scheduling stage begins. Under the dynamic operation environment of the production line, the inspection target selects an optimal inspection intervention point from all potential workstation gaps (such as idle time periods between processes) available for inspection task intervention. This ensures both the accurate execution of tensioner status inspection and minimizes interference with the main production process and other key inspection tasks.

[0101] Based on the real-time operating status of the production line (such as the current speed of the conveyor belt, the task queue of the robotic arm, and the cycle time of each process), the system automatically scans and extracts all potential workstation gaps that can be used for inspection tasks. Gaps are represented as: the idle time period between the completion of one process and the start of the next (e.g., the idle time window before a tensioner enters the inspection station after assembly); the interval area between adjacent products on the conveyor line (e.g., the time period corresponding to the minimum safe distance maintained by two tensioner assemblies on the conveyor belt); and the idle time period of inspection equipment (e.g., cameras, sensors) or actuators (e.g., the available time between the completion of the previous inspection task and the triggering of the next task).

[0102] Each potential workstation gap is marked as a candidate location, and its time range (start time t_start and end time t_end), spatial location (such as the coordinate range of the conveyor line), and associated equipment resources (such as the occupied detection camera number) are recorded.

[0103] For each candidate location, evaluate whether it conflicts with the existing detection task in terms of timing. A timing conflict is defined as:

[0104] The time range of candidate positions overlaps with the execution time range of a scheduled detection task (e.g., from the start time t_task_start to the end time t_task_end of the detection action) (i.e., [t_start, t_end] ∩ [t_task_start, t_task_end] ≠ If overlap exists, the candidate position is determined to have a temporal conflict with the corresponding detection task. For example:

[0105] If the time window for a candidate location is [10:00:00, 10:00:05], and the execution time for a scheduled detection task (such as defect detection of tensioner A) is [10:00:03, 10:00:07], then the two conflict within the time period [10:00:03, 10:00:05], and priority evaluation of this conflict is required.

[0106] For each detection task that conflicts with a candidate location in terms of timing, an avoidance priority index is calculated based on its criticality index in the production line and its overall cycle time impact index. This index quantifies the necessity for the task to be avoided. A higher avoidance priority index indicates a more critical task, and the candidate location should actively avoid this task to prevent impacting mainline production.

[0107] The Criticality Index (CI) reflects the impact of a testing task on product quality or production line safety. It is determined by extracting the type of the current testing task (such as bolt torque testing, sealing test, etc.), associating it with the specific testing object (such as the bolt assembly of a tensioner), querying the real-time quality database, and calculating the defect conversion rate (number of final product defects caused by the failure of this test / number of test failures) of this test item in the past M batches (such as the most recent 30 batches) as the Criticality Index. If the bolt torque test failed 10 times in the past 30 batches, and 8 of them caused the tensioner locking failure (defect), then CI = 8 / 10 = 0.8.

[0108] Overall Takt Time Impact Index (TI): This measures the cascading effect of delaying or canceling a testing task on the overall takt time (i.e., production efficiency) of the production line. It is calculated by taking the number of subsequent processes blocked by the testing task (e.g., assembly can only proceed after the tensioner testing is completed) and dividing the number of subsequent processes blocked by the testing task by the total number of processes on the production line to obtain the Overall Takt Time Impact Index (i.e., all processes for processing the tensioner, including production processes, testing processes, etc.). The higher the Overall Takt Time Impact Index, the higher the priority of avoiding the testing task.

[0109] The detection criticality index and the overall cycle time impact index are normalized using a maximum-minimum method. The normalized detection criticality index and the overall cycle time impact index are then summed to obtain the avoidance priority index (PI). The higher the PI value, the more priority the detection task needs to be ensured, and the more the corresponding candidate positions need to be avoided.

[0110] For each candidate location, the avoidance priority indices (PI_1, PI_2, ..., PI_m) of all conflicting detection tasks are summed, where m is the number of all conflicting detection tasks. The avoidance coefficient for that location is calculated by summing these indices. The avoidance coefficient reflects the overall avoidance pressure that the candidate location needs to bear due to temporal conflicts—the smaller the AC value, the lower the total priority of the conflicting tasks (i.e., the smaller the avoidance requirement), and the more suitable the candidate location is as a detection intervention point.

[0111] Finally, all candidate positions are traversed, and the position with the smallest avoidance coefficient AC is selected as the detection intervention point. At this position, a precise detection task of the tensioner is triggered (such as starting the camera to acquire images and calling the algorithm to analyze the tensioner status). This selection strategy ensures the efficient execution of the detection operation in the dynamic environment of the production line, meeting the accuracy requirements of tensioner detection while minimizing interference with the main production process and other critical detection tasks (such as avoiding delays in high-priority tasks or production line stoppages due to detection intervention).

[0112] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0113] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A tensioner detection method based on production line-product coupling, characterized in that: The detection method Includes the following steps: A1: Collect the content to be inspected, which covers multiple inspected targets, each of which consists of several feature elements; A2: Based on the spatial distribution characteristics of the feature elements in each inspected target and combined with the temporal entry information, preliminary clustering is performed on all feature elements, and feature elements belonging to the same inspected target are aggregated into a feature set to form a preliminary detection feature profile of the inspected target. A3: When the detection field of view of any two inspected targets overlaps, based on the spatial overlap analysis results and temporal entry information, the attribution judgment is carried out on the feature elements within the overlapping area. The comprehensive matching score is calculated based on the spatial proximity and temporal matching degree of the feature elements, and the feature elements are assigned to the inspected target with the highest comprehensive matching score. A4: When performing inspection operations on the target under inspection, automatically scan and analyze all potential workstation gaps on the current production line that can be used for inspection tasks as candidate positions. Analyze whether the candidate positions have timing conflicts with existing inspection tasks. If there is a conflict, obtain the avoidance coefficient of each candidate position and select the candidate position with the smallest avoidance coefficient as the inspection intervention point to inspect the target under inspection.

2. The tensioner detection method based on production line-product coupling according to claim 1, characterized in that: The attribution of feature elements within the overlapping area is determined, including the entry sequence of the two inspected targets and the relative positional relationship between the feature within the overlapping area and the visual boundaries of the two inspected targets.

3. The tensioner detection method based on production line-product coupling according to claim 2, characterized in that: Assigning feature elements to the tested target with the highest overall matching score includes the following steps: Based on the predefined detection field of view boundary, the overlapping area of ​​the fields of view of any two inspected targets is calculated, and the overlapping area is determined by geometric operations in coordinate space; Filter all inspected targets whose feature coordinates are located within the overlapping area to form a set of feature elements to be judged, S_overlap; Iterate through each feature element s_i in the feature set S_overlap, calculate its comprehensive matching score with all relevant inspected targets, and assign s_i to the inspected target with the highest comprehensive matching score.

4. The tensioner detection method based on production line-product coupling according to claim 3, characterized in that: The spatial proximity is the Euclidean distance between the feature element and the center of the detection field of the subject being inspected; the smaller the distance, the stronger the spatial correlation. The temporal matching degree is the degree to which the timing of the appearance of a feature element matches the timing of the entry of a certain inspected target. The higher the matching degree, the closer the temporal correlation.

5. The tensioner detection method based on production line-product coupling according to claim 4, characterized in that: The comprehensive matching score is calculated based on the spatial proximity and temporal matching degree of the feature elements, including the following steps: the spatial proximity and temporal matching degree are respectively subjected to maximum-minimum normalization processing, and the normalized spatial proximity and temporal matching degree are weighted to calculate the comprehensive matching score between each feature element and each subject. Among them, spatial proximity is an inverse proportional factor, and temporal matching degree is a direct proportional factor.

6. The tensioner detection method based on production line-product coupling according to claim 5, characterized in that: The spatial proximity is used to quantify the spatial correlation strength between a feature element and the center of the detection field of a certain inspected target. For each inspected target, its center coordinates are calculated through the geometric center of all clustered feature elements of the inspected target. For each feature element s_i in the feature element set S_overlap, the Euclidean distance d_ij between it and the center of the field of view C_j of the inspected target T_j is calculated as the spatial proximity.

7. The tensioner detection method based on production line-product coupling according to claim 5, characterized in that: The time-series matching degree is used to quantify the temporal correlation between the occurrence time of a feature element and the entry time of a certain inspected target. The entry time t_Tj of each inspected target is determined by its actual arrival timestamp on the production line. The time sequence t_si of feature element s_i is associated with its original detection data stream. For each feature element s_i in the feature element set S_overlap, the absolute deviation Δt_ij between its time sequence t_si and the entry time t_Tj of the inspected target T_j is calculated, and then the absolute deviation Δt_ij is converted into the time-series matching degree.

8. The tensioner detection method based on production line-product coupling according to claim 1, characterized in that: Obtain the avoidance coefficient for each candidate position. The calculation logic for the avoidance coefficient is as follows: Based on the detection criticality index of the detection task in the production line and the overall cycle time impact index, an avoidance priority index is generated. The larger the avoidance priority index, the more the candidate position needs to avoid the current detection task. The avoidance coefficient is obtained by summing the avoidance priority indices of all detection tasks at each candidate position.

9. The tensioner detection method based on production line-product coupling according to claim 8, characterized in that: When performing inspection operations on the target, the system automatically scans and analyzes all potential workstation gaps on the current production line that are available for inspection tasks, as candidate positions. It then analyzes whether these candidate positions conflict with existing inspection tasks in terms of timing. If a conflict exists, it obtains the avoidance coefficient for each candidate position and selects the candidate position with the smallest avoidance coefficient as the inspection intervention point to inspect the target. This includes the following steps: If the time range of a candidate location overlaps with the execution time range of a scheduled detection task, it is determined that the candidate location and the corresponding detection task have a timing conflict. For each detection task that has a timing conflict with the candidate position, the avoidance priority index of the detection task is calculated based on its criticality index in the production line and the overall cycle time impact index. The detection criticality index and the overall beat impact index are normalized by the maximum and minimum values. The avoidance priority index is obtained by summing the normalized detection criticality index and the overall beat impact index. For each candidate location, the avoidance priority index of all conflicting detection tasks is aggregated, and the avoidance coefficient of the candidate location is calculated by summing the avoidance priority indices of all detection tasks.

10. A tensioner detection system based on production line-product coupling, used to implement the detection method according to any one of claims 1-9, characterized in that: It includes an overlapping area analysis module, an attribution determination module, and an avoidance analysis module; Overlapping Area Analysis Module: Collects the content to be inspected, which covers multiple inspected targets. Each inspected target is composed of several feature elements. Based on the spatial distribution characteristics of the feature elements in each inspected target and combined with the temporal entry information, preliminary clustering is performed on all feature elements. Feature elements belonging to the same inspected target are aggregated into a feature set to form a preliminary detection feature profile of the inspected target. When the detection field of view of any two inspected targets overlaps, a discrimination criterion is introduced to analyze the inspected target affiliation relationship of the feature elements in the overlapping area. Attribution determination module: For overlapping areas, based on spatial overlap analysis results and temporal entry information, attribution determination is carried out on feature elements within the overlapping area. The comprehensive matching score is calculated based on the spatial proximity and temporal matching degree of the feature elements, and the feature elements are assigned to the inspected target with the highest comprehensive matching score. Avoidance Analysis Module: When performing inspection operations on the target under inspection, it automatically scans and analyzes all potential workstation gaps on the current production line that can be used for inspection tasks as candidate positions. It analyzes whether the candidate positions have timing conflicts with existing inspection tasks. If there is a conflict, it obtains the avoidance coefficient of each candidate position and selects the candidate position with the smallest avoidance coefficient as the inspection intervention point to inspect the target under inspection.

Images