Machine vision-based meshing precision measurement and calibration method for gear transmission system

Abstract

The invention relates to a machine vision-based meshing precision measurement calibration method for a gear transmission system, and belongs to the technical field of calibration workpieces. The method comprises the following steps: extracting a gradient direction of a contact edge point in an image, analyzing a direction change trend, removing a mutation point, and constructing an edge direction paragraph sequence; positioning a tooth profile contour intersection point according to the sequence, analyzing longitudinal coordinate fluctuation, and identifying an axial dislocation structure; comparing the actual tooth crest contour with a standard path, extracting the direction and length of a difference section, and generating an adjustment direction mark group; converting the direction mark into a fine adjustment angle sequence, collecting an image and identifying a direction stable section; and finally, dividing image sub-blocks, analyzing the curvature direction of a spliced boundary, identifying curvature mutation points and establishing a calibration scheme. According to the method, the integrity of gear meshing area structure recognition and the continuity of an image sequence are effectively improved, high-sensitivity axial dislocation detection and accurate direction guiding fine adjustment are achieved, the curvature sudden change recognition and multi-dimensional error response capability is enhanced, the engineering practicability and the automation level are high, and the method is suitable for large-scale popularization and application. The method is suitable for metering calibration and quality control of a high-precision gear transmission system.

Description

Technical Field

[0001] This invention relates to the field of workpiece calibration technology, and in particular to a machine vision-based method for measuring and calibrating the meshing accuracy of gear transmission systems. Background Technology

[0002] As a core component in the field of mechanical transmission, the meshing accuracy of gear transmission systems directly affects transmission efficiency, noise levels, and service life. Traditional methods for measuring gear meshing accuracy typically rely on contact measurements or visual methods based on static image analysis, which have many limitations. Especially during dynamic meshing, due to factors such as gear manufacturing errors, assembly deviations, or wear, the contact area of ​​the tooth surface is prone to microscopic deformation, positional drift, or edge jumps, making it difficult for traditional methods to achieve high-precision, continuous error detection and calibration.

[0003] Existing machine vision-based methods for measuring gear meshing accuracy, while achieving non-contact measurement to some extent, still have the following technical problems: Poor structural coherence in image sequences makes stable edge tracking difficult: Traditional methods, when processing multi-frame images, often rely on static edge feature recognition and lack the ability to model the consistency of edge direction changes between frames. Under conditions of vibration, drift, or illumination changes, the direction of edge points is prone to abrupt changes, leading to path breaks or mismatches, making it impossible to construct a complete structural motion trajectory, and even more difficult to achieve continuous capture and analysis of dynamic errors.

[0004] Low sensitivity of axial misalignment detection and lag in position fluctuation response: Due to the lack of longitudinal fluctuation density analysis of the tooth profile intersection path, traditional methods struggle to distinguish between normal meshing motion and abnormal axial misalignment. Especially when the misalignment amplitude is small or in its initial stage, it is impossible to identify the fluctuation accumulation area in a timely manner, resulting in lag in misalignment detection and delay in calibration.

[0005] The lack of directional markings for the difference segments leads to a lack of guidance in the fine-tuning strategy: When comparing the actual tooth profile with the standard path, existing methods mostly focus on calculating the distance difference, without systematically extracting the directional features and coverage of the difference segments, resulting in the inability to form a structured set of directional markings. Consequently, they cannot provide clear and discrete directional guidance for subsequent angle fine-tuning, affecting the systematic nature and accuracy of the calibration.

[0006] The traditional methods suffer from weak curvature abrupt change interface recognition capabilities and poor consistency in sub-block stitching structures. In image sub-block stitching and boundary analysis, the lack of a mechanism to determine the continuity of curvature direction makes it difficult to detect abrupt slope changes and their specific locations within the structure. This results in the calibration scheme being unable to accurately locate interface errors, and the inability to trace and compensate for stitching deviations between sub-blocks, thus limiting the improvement of overall calibration accuracy.

[0007] The calibration scheme lacks multi-dimensional error integration and dynamic response mechanisms: existing methods typically stop at the error detection stage, failing to integrate multi-dimensional features such as edge direction, positional fluctuations, and curvature changes into an executable calibration scheme, and further lacking the ability to judge stability and provide feedback during dynamic adjustment. This results in the calibration process relying on manual intervention, having a low degree of automation, and being unable to cope with the accuracy maintenance requirements under complex working conditions.

[0008] Therefore, it is necessary to propose a gear meshing accuracy measurement and calibration method that can integrate edge direction analysis, fluctuation density judgment, difference direction marking, curvature change identification and dynamic response mechanism, so as to improve the continuity of detection, misalignment sensitivity, direction guidance, structural integrity and calibration automation level. Summary of the Invention

[0009] To address the technical problems existing in the prior art, embodiments of the present invention provide a machine vision-based method for measuring and calibrating the meshing accuracy of a gear transmission system, comprising the following steps: To achieve the above objectives, the present invention adopts the following technical solution: a machine vision-based method for measuring and calibrating the meshing accuracy of a gear transmission system, comprising the following steps: S1: Extract contact edge points in the image, calculate the gradient direction of adjacent pixels, mark the angle direction, analyze the trend of direction change, remove abrupt points, construct direction segments and inter-frame indexes according to frame order, and generate edge direction segment sequence. S2: Based on the edge direction segment sequence, locate the intersection of the tooth profile, extract the longitudinal coordinates, analyze the changes in the path and distance of the intersection points within the frame, remove the sequences with no changes, evaluate the remaining fluctuation density, locate the tooth axis, and obtain the axial misalignment structure. S3: Based on the axial misalignment structure, extract the tooth tip profile curve, compare it with the standard path, calculate the point-to-point spatial difference, extract the direction and length of the difference segment, organize the regional structure and mark the discrete direction to form an adjustment direction mark group; S4: Based on the adjustment direction marker group, the direction markers are converted into a fine-tuning angle sequence, image frames are acquired and the direction changes are compared, stable segments are identified according to the minimum rate threshold, and adjustment direction segments are generated. S5: Based on the adjustment direction segment, extract the curvature direction of the image sub-block splicing boundary, analyze the direction continuity, identify the slope change point, trace back the path to locate the sub-block index, and establish a meshing accuracy measurement and calibration scheme.

[0010] As a further aspect of the present invention, the edge direction segment sequence includes edge point direction angles, continuous direction segment numbers, inter-frame correspondence, and structural path indexes; the axial misalignment structure includes tooth intersection coordinates, frame sequence distribution trajectory, position fluctuation amplitude, and misalignment positioning segments; the adjustment direction marker group includes boundary difference segments, direction change, length coverage, and discrete direction angle identifiers; the adjustment direction segment includes angle step sequence, direction change rate, stable direction segment, and continuity verification results; and the meshing accuracy measurement and calibration scheme includes sub-block splicing numbers, curvature change paths, abrupt connection point indexes, and interface marking information.

[0011] As a further aspect of the present invention, the unchanged sequence refers to a continuous data sequence in which the path and distance of the intersection points within a frame remain constant.

[0012] As a further aspect of the present invention, the minimum rate threshold refers to the minimum rate threshold used to determine the stability of directional change.

[0013] As a further aspect of the present invention, the specific steps of S1 are as follows: S101: Obtain the grayscale information of the gear meshing image, and use the edge detection operator to extract the set of edge points of the workpiece contact area for each frame of the image. Construct the edge point sequence according to the position information of the edge points in the image coordinates, calculate the grayscale gradient direction of adjacent pixels and mark it with angles, and generate the set of edge gradient direction angles. S102: Call the edge gradient direction angle set, perform directional change trend analysis on continuous angle direction segments, calculate the difference between adjacent angles and compare it with the edge direction change amplitude threshold, remove angle points with differences greater than the edge direction change amplitude threshold, and obtain the continuous edge direction segment set. S103: Based on the set of continuous segments along the edge direction, perform inter-frame angle fitting degree matching on the edge segments, calculate the average direction difference and compare it with the inter-frame direction fitting degree threshold, establish a matching index relationship and remove duplicate segments to generate an edge direction segment sequence.

[0014] As a further aspect of the present invention, the specific steps of S2 are as follows: S201: Based on the edge direction segment sequence, determine the direction intersection of the tooth-shaped contours in the image, locate the position of the intersection in the image coordinates, extract the coordinate values ​​in the vertical direction, and generate a set of tooth-shaped intersection position information. S202: Call the tooth-shaped intersection point location information set, calculate the longitudinal coordinate difference of the intersection points in the frame sequence, establish the intersection point distribution path, compare the distance change trend between adjacent points, filter out the path segments that have not undergone structural movement, and obtain the structural fluctuation distribution path set; S203: Based on the set of structural fluctuation distribution paths, extract the longitudinal position change values ​​in the path segments, determine the density according to the fluctuation interval, filter the fluctuation path segments and locate the coordinate positions to obtain the axial misalignment structure.

[0015] As a further aspect of the present invention, the specific steps of S3 are as follows: S301: Based on the axial misalignment structure, extract continuous edge point paths in the tooth tip contour of the workpiece, connect them in pixel order to form a closed curve, establish the boundary trajectory in the workpiece image, and call the design standard contact path coordinate group to perform point-by-point spatial distance calculation on the two curves in the image plane to obtain the boundary curve difference fragment set. S302: Based on the boundary curve difference segment set, extract the path direction change data and the total number of pixels corresponding to each difference segment, calculate the direction angle based on the coordinate sequence direction change, and count the coverage length range to obtain the difference path direction coverage set; S303: Call the differential path direction coverage set, divide the path direction into several discrete direction categories according to the direction angle interval, aggregate the marked direction data according to the region, establish the distribution position of the direction segment in the image coordinates and the relationship of the assigned direction identifier, and obtain the adjusted direction mark group.

[0016] As a further aspect of the present invention, the specific steps of S4 are as follows: S401: According to the adjustment direction mark group, the direction marks are converted into an angle step sequence and input into the rotation adjustment platform. Image frames are acquired at each angle step, and edge extraction and direction sequence reconstruction are performed on the image to obtain the rotation step edge structure set. S402: Call the rotation stepping edge structure set, compare the boundary direction sequence in the image, calculate the rate of change of the angle between the direction vectors of adjacent frames, and locate the region with the slowest direction change based on the minimum rate threshold to obtain the image direction stable interval group. S403: Based on the image direction stable interval group, determine whether the image frame sequence in each region meets the angle step continuity condition, filter out all complete region segments of the sequence, extract the angle start and end range, and generate the adjustment direction segment.

[0017] As a further aspect of the present invention, the specific steps of S5 are as follows: S501: According to the adjustment direction segment, divide the image into sub-blocks in the corresponding image frame, extract the edge direction on the sub-block splicing boundary, calculate the boundary direction curvature value, make a continuity judgment on the curvature direction change value between adjacent sub-blocks, and establish a splicing boundary curvature direction set. S502: Call the set of curvature directions of the splicing boundary, construct the curvature change path map in sequence in the splicing path of the image sub-blocks, calculate the curvature slope change value of adjacent boundary segments and compare it with the curvature change judgment threshold, identify the position corresponding to the change slope point, and obtain the curvature change point index set. S503: Based on the curvature mutation point index set, retrieve the corresponding splicing path number and perform path direction backtracking positioning, extract the index number of adjacent image sub-blocks at the mutation point and mark them, aggregate all marking results, and establish a meshing accuracy measurement and calibration scheme.

[0018] As a further aspect of the present invention, the curvature abrupt change determination threshold refers to the minimum change amplitude critical value used to identify whether the change in curvature slope of adjacent boundary segments is abnormal.

[0019] Compared with the prior art, the advantages and positive effects of the present invention are as follows: 1. Improve the completeness and consistency of structural recognition: By extracting the gradient directions of contact edge points in the image, the system analyzes the trend of direction changes and removes abrupt changes to construct a continuous and stable sequence of edge direction segments. This method overcomes the problem of traditional methods struggling to maintain boundary direction consistency under dynamic disturbances or misalignments, ensuring the coherent construction of structural paths across multiple frames and significantly improving the recognition integrity and tracking stability of edge structures in the gear meshing region.

[0020] 2. Precisely locate axial misalignment and structural fluctuations: Based on the edge-direction segment sequence, the system further locates the intersection points of the tooth profile, analyzes their longitudinal coordinate fluctuations, identifies unchanged sequences, and evaluates the fluctuation density, thereby accurately determining the axial misalignment region. This method can effectively distinguish between normal motion and abnormal misalignment, achieving highly sensitive detection and localization of minute positional drifts during gear meshing.

[0021] 3. Efficiently generate adjustment direction markers and fine-tuning strategies: By comparing the spatial differences between the actual tooth tip profile and the standard path, the direction and length of the difference segments are extracted to form a discrete direction marker group, which is then converted into an angle step sequence. This method can guide the rotary adjustment platform to perform precise fine-tuning. Combined with image acquisition and direction change rate analysis, it can identify directionally stable segments, significantly improving the automation and accuracy of the adjustment process.

[0022] 4. Enhanced curvature change recognition and interface mapping capabilities: Within the adjustment direction segment, the system further divides the image into sub-blocks, analyzes the continuity of curvature direction at the splicing boundaries, identifies abrupt slope changes, and precisely locates the sub-block number where the abrupt interface is located through path backtracking. This method can effectively detect structural discontinuities during meshing, establish a complete curvature change path and interface marking information, and provide a clear structural basis for subsequent calibration.

[0023] 5. Supports multi-dimensional error response and calibration scheme generation: The final meshing accuracy calibration scheme encompasses multi-dimensional information such as sub-block assembly numbers, curvature change paths, and abrupt connection point indices, possessing strong error localization and response capabilities. This method is not only applicable to static error calibration but also provides real-time feedback on directional stability during dynamic adjustment, significantly improving the overall metrological accuracy and reliability of the gear transmission system.

[0024] 6. High adaptability and high engineering practical value: This method introduces multiple adaptive thresholds (such as the edge direction change amplitude threshold and curvature change judgment threshold) into image processing, and sets parameters based on actual image resolution and working conditions, exhibiting strong engineering adaptability and practicality. It can be widely applied in high-precision gear manufacturing, assembly, debugging, and operation and maintenance, providing reliable technical support for quality control and performance optimization of gear transmission systems.

[0025] In summary, this invention, by integrating machine vision and image processing technologies, achieves high-precision extraction and analysis of multi-dimensional features such as edge direction, position fluctuation, and curvature change during gear meshing, significantly improving the automation, accuracy, and reliability of the calibration process, and has significant engineering application value and promotion prospects. Attached Figure Description

[0026] 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.

[0027] Figure 1 This is a schematic diagram of the steps of the present invention; Figure 2 This is a detailed schematic diagram of S1 of the present invention; Figure 3 This is a detailed schematic diagram of S2 of the present invention; Figure 4 This is a detailed schematic diagram of S3 of the present invention; Figure 5 This is a detailed schematic diagram of S4 of the present invention; Figure 6 This is a detailed schematic diagram of S5 of the present invention. Detailed Implementation

[0028] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0029] 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.

[0030] 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, their intended meanings are consistent. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent.

[0031] 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.

[0032] 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.

[0033] Please see Figure 1 This invention provides a machine vision-based method for measuring and calibrating the meshing accuracy of a gear transmission system, comprising the following steps: S1: Obtain the set of edge points of the workpiece contact area in the gear meshing image, extract the gradient direction of adjacent pixels in each set of edge point sequences and mark the angle direction, perform difference analysis on the boundary direction change trend in the continuous direction and remove the turning amplitude abrupt point, sort the stable region according to the image frame order and generate continuous direction segments, construct the inter-frame correspondence index relationship between edge directions of multiple frames and remove duplicate segments, connect the stable direction segments and generate the structure index path set, and obtain the edge direction segment sequence; S2: Based on the edge direction segment sequence, locate the intersection points of the workpiece tooth profiles and extract the position coordinates in the vertical direction of the image. Extract the distribution path of the intersection points in the frame sequence and compare the trend of the distance change between the points. Filter out sequences with no structural movement changes between adjacent intersection points. Analyze the position fluctuation values ​​in the residual sequence and judge the density according to the change range. Select the frequently fluctuating segments and locate the positions distributed on the tooth axis to obtain the axial misalignment structure. S3: Based on the axial misalignment structure, extract the continuous edge point path in the tooth tip contour of the workpiece and connect it into an outer contour curve. Read the design standard contact path coordinate group and compare it with the boundary trajectory in the workpiece image. Perform point-to-point spatial distance comparison on the two curves in the image plane, screen out the segments with significant differences and extract the direction and length coverage. Organize them into a regional distribution structure and divide them into discrete direction marks to form an adjustment direction mark group. S4: Based on the adjustment direction mark group, each direction mark is converted into an angle fine-tuning sequence and applied to the rotation adjustment platform. Image frames under the adjustment angle are acquired and the edge direction structure is re-extracted. The boundary direction sequences in the images between continuous adjustment steps are compared. The image direction stable region is determined according to the principle of minimum direction change rate. The rotation interval range that meets the sequence continuity condition is screened out to generate the adjustment direction segment. S5: Based on the adjustment direction segment, perform splicing boundary curvature direction extraction operation on the image sub-blocks within the adjustment segment, determine the directional continuity between adjacent sub-block splicing boundaries and construct a curvature change path map, identify the sub-block connection points where the curvature change slope changes abruptly in the splicing sequence, locate the structural splicing path number before and after the abrupt change point and perform backtracking mapping, extract the workpiece image sub-block index number at the abrupt change interface and generate a mark, and establish a meshing accuracy measurement and calibration scheme.

[0034] The edge direction segment sequence includes edge point direction angle, continuous direction segment number, inter-frame correspondence, and structural path index. The axial misalignment structure includes tooth intersection coordinates, frame sequence distribution trajectory, position fluctuation amplitude, and misalignment positioning segment. The adjustment direction mark group includes boundary difference segment, direction change, length coverage, and discrete direction angle identifier. The adjustment direction segment includes angle step sequence, direction change rate, stable direction segment, and continuity verification results. The meshing accuracy measurement and calibration scheme includes sub-block splicing number, curvature change path, abrupt connection point index, and interface mark information.

[0035] Please see Figure 2 The specific steps of S1 are as follows: S101: Obtain the grayscale information of the gear meshing image, and use the edge detection operator to extract the set of edge points of the workpiece contact area for each frame of the image. Construct the edge point sequence according to the position information of the edge points in the image coordinates, calculate the grayscale gradient direction of adjacent pixels and mark it with angles, and generate the set of edge gradient direction angles. The acquisition of image grayscale information is based on image acquisition of the gear meshing area. Typically, a high-resolution industrial camera is used to capture images frame-by-frame while the gear is rotating. Each frame consists of RGB channels. To facilitate subsequent processing, these images need to be converted to grayscale. During the process, each pixel in the image is traversed, and its R, G, and B channel values ​​are read. A single grayscale value is generated through weighted aggregation. For example, for a pixel (R=120, G=110, B=100), the calculated grayscale value is approximately 109. This value is used as the grayscale value of that pixel in the new image. After completing the grayscale conversion of the entire image, edge detection is performed on the grayscale image. A common method is to define a rectangular region in the image to define the meshing area. For each pixel within the defined region, the grayscale values ​​of several surrounding pixels are read, and the difference is calculated. Areas with significant grayscale changes are identified as edge points, and their positions are recorded using image coordinates after detection. For example, if an edge point is detected at... The image coordinates (200, 180) are recorded and added to the edge point set. Simultaneously, the direction of the grayscale difference between this point and its neighboring points is calculated. By comparing the gradient direction of adjacent pixel grayscale values, the specific direction angle information is marked for this point. For example, if the grayscale of a pixel to the left of an edge point is 100 and to the right is 130, it indicates that the grayscale change direction is from left to right, and its angle direction is marked as horizontal to the right. The same operation is performed for each edge point, storing the direction angle in the set. Then, the edge points are sorted based on the image coordinates to generate a continuous edge point sequence. For example, three adjacent edge points located at (100, 150), (101, 150), and (102, 150) are arranged in sequence to form an edge point sequence. Combined with their corresponding angle information, an edge gradient direction angle set is formed. This process is performed in all frames of the image, building and saving the angle set frame by frame for subsequent angle direction consistency analysis.

[0036] S102: Call the edge gradient direction angle set, perform directional change trend analysis on continuous angle direction segments, calculate the difference between adjacent angles and compare it with the edge direction change amplitude threshold, remove angle points with differences greater than the edge direction change amplitude threshold, and obtain the continuous edge direction segment set. After obtaining the edge gradient direction angle set for each frame of the image, the direction angle values ​​stored in the angle set are called one by one, and the continuous angle data in each edge point sequence are traversed. For each pair of adjacent angles, the difference between them is calculated to determine whether a sudden change in direction has occurred. The calculation method is to first read the i-th angle value and the (i+1)-th angle value in the sequence, and perform the difference calculation between the angle values. If the absolute value of the difference exceeds the threshold of the edge direction change amplitude, it is determined to be a direction change point and is removed. The remaining segment angle set with relatively consistent directions is retained as a continuous segment set. The threshold setting needs to consider the normal angle change range that may occur in the actual image edge direction. It is usually set by sample statistics method, by analyzing multiple images of similar workpieces. Each edge point segment is sampled, and the average and standard deviation of adjacent angle changes are calculated. For example, after sampling 500 angle segments, it is found that most of the angle differences are concentrated between 3° and 8°. Based on the standard deviation calculation results, the threshold is set to 10°, which means that points with angle differences greater than 10° will be removed and no longer considered as continuous direction segments. If a certain angle sequence is [35°, 36°, 37°, 54°, 55°], after performing difference analysis, it is found that the difference between 37° and 54° is 17°, which exceeds the set threshold. Therefore, 54° and its subsequent points are removed, and the first three points with consistent directions are retained to generate continuous segments. This process is repeated in all image frames to finally obtain a set of continuous edge direction segments, which serves as the data source for subsequent matching analysis.

[0037] S103: Based on the set of continuous segments in the edge direction, perform inter-frame angle fitting degree matching on the edge segments, calculate the average direction difference and compare it with the inter-frame direction fitting degree threshold, establish a matching index relationship and remove duplicate segments to generate an edge direction segment sequence. Inter-frame analysis is performed on the set of continuous edge direction segments obtained in the previous processing step. This requires establishing the correspondence between angle segments between frames. First, the set of direction segments of the current frame and the adjacent next frame is read. Then, a segment from the current frame is selected and matched with all segments in the next frame to perform an average direction angle matching operation. For each segment, the average value of all angle values ​​within that segment is calculated and compared with the average value of each segment in the target frame. For example, if the current frame segment angle is [42°, 43°, 44°] with an average angle of 43°, and a segment in the target frame has angles [45°, 46°, 47°] with an average angle of 46°, the difference is 3°. If the preset inter-frame direction fitting threshold is 5°, then 3° is less than the threshold and is considered a good fit. The process involves establishing an index relationship between segments. If multiple matching segments appear in the target frame for the current segment, the segment with the smallest difference is selected for pairing, and the remaining segments are discarded to avoid redundant or repeated segment matching. The inter-frame orientation fitting threshold is usually set based on the actual directional changes and fluctuations of gear meshing images between different frames in engineering. The angle difference of the real matching segments is extracted from multiple sets of continuous frame samples using a measured statistical method, and the 95th percentile is calculated as the judgment criterion. For example, in 500 sets of real inter-frame segment pairings, 475 sets have an average difference between 0° and 4.5°, so the inter-frame fitting threshold can be set to 5° for judging the upper and lower limits of actual segment matching. The entire process is executed segment by segment in all image frames, and finally, an edge orientation segment sequence is generated.

[0038] Please see Figure 3 The specific steps of S2 are as follows: S201: Based on the edge direction segment sequence, determine the direction intersection of the contours between the teeth in the image, locate the position of the intersection in the image coordinates, extract the coordinate values ​​in the vertical direction, and generate a set of tooth intersection position information. First, the established edge direction segments in each frame of the image are read, and the angular features of adjacent direction segments are analyzed one by one to determine whether there is an angular intersection relationship. Specifically, the start and end point direction values ​​of each pair of direction segments are extracted, and the intersection relationship is determined by comparing whether the two direction segments form a certain angular difference. For example, when one direction is horizontal to the right (angle approximately 0°) and the adjacent direction is vertical upward (angle approximately 90°), the angle between the two direction segments is close to 90°, indicating an intersection trend. For direction segment pairs that meet the intersection trend, the two direction segments are extended linearly according to the image coordinate system, and the intersection point of the two extensions is recorded as the intersection point of the tooth-shaped contour. Then, the specific coordinate position of the intersection point in the image is read. The vertical coordinate value, i.e., the Y-axis coordinate, is extracted and used as the vertical positioning data. The image frame number to which the intersection point belongs and its corresponding horizontal and vertical coordinate information are recorded. Finally, the intersection point position information set is constructed. In actual images, for the tooth-shaped contour area, 8 to 12 sets of intersection points can usually be identified per frame. In actual engineering, taking an image resolution of 1920×1080 as an example, the vertical coordinate value of the intersection point usually falls between 300 and 800 pixels. To ensure the integrity of this set of data, the image frame number, intersection point number, horizontal coordinate, vertical coordinate, and corresponding edge segment information of each intersection point are recorded frame by frame and stored in a structured array according to the time series to support subsequent dynamic analysis operations and generate the tooth-shaped intersection point position information set.

[0039] S202: Call the tooth-shaped intersection point location information set, calculate the longitudinal coordinate difference of the intersection points in the frame sequence, establish the intersection point distribution path, compare the distance change trend between adjacent points, filter out the path segments that have not undergone structural movement, and obtain the structural fluctuation distribution path set; After retrieving the tooth-shaped intersection point location information set, a frame-by-frame extraction operation is performed on the vertical coordinate values ​​of each identified intersection point number in consecutive image frames. The vertical coordinates of the corresponding intersection points in each frame are arranged sequentially to form a sequence of vertical position changes. Then, the coordinate differences between adjacent frames in this sequence are calculated to obtain the vertical position change of the intersection point between frames. For example, the vertical coordinates of intersection point number 5 in frames 1 to 4 are 400, 402, 405, and 407, respectively, so its change is +2, +3, and +2. The same operation is performed on all intersection points in sequence to form complete inter-frame vertical change data. Based on this data, the coordinates of the same intersection point in each frame are connected to form a distribution path. The vertical position of each point in the path is the Y value, and the lines connecting the points form the intersection point movement trend track. After establishing the path, the longitudinal change trend of any two adjacent points in the path is further compared to determine whether it has obvious change characteristics. When performing the judgment, a structural movement judgment threshold needs to be set. If the longitudinal difference between two consecutive points is less than the threshold, it is determined that no structural movement has occurred in the path segment and it is removed. The threshold setting can be set according to the accuracy and actual equipment motion fluctuation measurement experience. For example, in industrial scenarios, image jitter or error is within ±1 pixel, so the threshold is set to 2 pixels. That is, when the difference of the Y value between two adjacent points is 0 or 1, it is considered that there is no movement, and such path segments are removed. Finally, segments with a difference of more than 2 pixels are retained to construct a structural fluctuation distribution path set. The path set is indexed by the intersection number and includes the start and end frame numbers and the corresponding Y value change sequence of each path segment.

[0040] S203: Based on the set of structural fluctuation distribution paths, extract the longitudinal position change values ​​in the path segments, determine the density according to the fluctuation interval, filter the fluctuation path segments and locate the coordinate positions to obtain the axial misalignment structure. The vertical coordinate sequence contained in each path is read one by one, and the coordinate differences between consecutive frames are organized to obtain the vertical change value sequence of each path between different frames. For example, if the vertical coordinates of a certain path are 410, 414, 417, 413, and 411, the differences are +4, +3, -4, and -2. Position change analysis is performed on the path to extract the vertical displacement of each frame segment, and all change values ​​are used as the basis for subsequent judgment. The fluctuation range operation is performed on the above change values, and different fluctuation level ranges are set. For example, within ±2 pixels is defined as weak fluctuation, within ±5 pixels as moderate fluctuation, and more than ±5 pixels as severe fluctuation. For each path, its fluctuation range is statistically analyzed. The number of frames within the interval is determined, and the density of the fluctuation segments is judged. For example, if a path has 6 frames with change values ​​falling within the range of ±5 or more in 10 frames, the fluctuation density is 60%, and it is marked as a high fluctuation path. Then, a fluctuation density screening threshold is set as the criterion for whether to enter the subsequent positioning. This threshold is set according to the actual working condition gear misalignment characteristics. If the fluctuation density of misaligned paths usually exceeds 50% in experience, then the threshold is set to 50%, that is, paths with a density greater than or equal to 50% will be marked as fluctuation segments. Then, the frame number and vertical position coordinates in this type of path are extracted, and the specific image frame and coordinate point are located according to the mapping relationship between the path number and the frame sequence, which is the location of the axial misalignment structure.

[0041] Please see Figure 4 The specific steps of S3 are as follows: S301: Based on the axial misalignment structure, continuous edge point paths are extracted from the tooth tip contour of the workpiece, and connected in pixel order to form a closed curve. The boundary trajectory in the workpiece image is established, and the design standard contact path coordinate group is called to perform point-by-point spatial distance calculation on the two curves in the image plane to obtain the boundary curve difference fragment set. Based on the axial misalignment structure, the misalignment area needs to be marked in the image frame and its corresponding workpiece tooth tip contour information needs to be extracted. During execution, the marked misalignment segment in the image coordinate system is used as a reference. The generated edge point data is called, and continuous edge pixel paths are retrieved in the tooth tip region. A path list is constructed according to the horizontal and vertical coordinates of each point, and it is determined whether the distance between adjacent pixels is a unit distance or within the set pixel continuity tolerance. The pixel continuity tolerance can be set to 1 pixel to ensure that there are no breaks or jumps in the path. When the distance between the first and last points of the entire path is less than the closure judgment threshold, the path is considered to form a closed curve. The closure judgment threshold is set according to the tooth tip contour size and image resolution, usually set to 3 pixels. If the pixel distance between the starting point and the ending point does not exceed 3 pixels, it is determined to be a closed path. After the closed curve is constructed, it is recorded as... The workpiece boundary trajectory curve is then imported, followed by the design standard contact path coordinate set. This coordinate set is standard tooth tip boundary data generated in advance based on CAD design. Each set of coordinate values ​​is a set of ideal position points at the image scale. After importing, point-by-point pairing is performed, comparing the coordinates of each point in the workpiece boundary trajectory with the corresponding point in the standard path. The image coordinates of each point are read and the actual pixel distance between the two points is calculated. The comparison results are judged against the set boundary error tolerance. The boundary error tolerance is set with reference to the tooth tip machining accuracy, which is usually 2 pixels at industrial-grade image resolution. If the distance of any point exceeds 2 pixels, it is considered an error point. The path segment where the error point is located is divided into boundary difference segments. This judgment process is repeated to form a set of boundary curve difference segments. Each segment record contains parameter information such as start coordinates, end coordinates, pixel distance difference, and image frame number.

[0042] S302: Based on the boundary curve difference segment set, extract the path direction change data and the total number of pixels corresponding to each difference segment, calculate the direction angle based on the coordinate sequence direction change, and count the coverage length range to obtain the difference path direction coverage set; Based on the boundary curve difference segment set, segment data is read segment by segment. For each difference segment, the coordinates of all pixels on the path are extracted and the number of points is counted to obtain the total number of pixels. For example, if a segment contains a total of 32 consecutive pixels from the start point to the end point, its total number of pixels is 32. Then, the coordinate sequence in the path is read in pixel order. For each pair of adjacent pixel coordinates, the difference between horizontal and vertical displacements is compared. By accumulating the change in the direction of the coordinate difference point by point, the overall direction of the path is determined, forming a direction change sequence. Based on this sequence, the average direction trend of the path segment is analyzed. For example, if the coordinate changes between consecutive points are mainly concentrated in the positive x-axis direction, the overall direction of the path segment is determined to be biased towards the horizontal direction. If there is a sudden change in angle in the direction sequence, such as a sudden change from horizontal to vertical, it is determined to be a sudden change in direction, and a sudden change in direction angle judgment is set. A threshold is used to distinguish whether a path belongs to a single direction. The threshold for abrupt changes in direction angle is generally set at 15 degrees, which is determined based on the path curvature characteristics and edge changes in the image. When the change angle between adjacent directions exceeds 15 degrees, it is judged as a change in direction. The segments with changes in direction are segmented according to the location of the change point, and separated into multiple sub-segments with the same direction. Then, the coverage length range of each segment is calculated. The coverage length is the continuous path length from the start point to the end point on the image plane. This length value is obtained by successively accumulating the distance between adjacent pixels. For example, the distance between each pair of adjacent points is about 1.4 pixels. If a segment consists of 10 points, the coverage length is about 13 pixels. The direction angle, number of pixels, and coverage length of each segment are combined and stored in the differential path direction coverage set. Each record corresponds to a set of direction information and its spatial extension information in the image.

[0043] S303: Call the differential path direction coverage set, divide the path direction into several discrete direction categories according to the direction angle interval, aggregate the marked direction data according to the region, establish the distribution position of the direction segment in the image coordinates and the relationship of the assigned direction label, and obtain the adjusted direction label group; After calling the differential path direction coverage set, a classification operation is performed on the direction angle values ​​in each record. The path direction is discretized according to a preset direction interval division standard. Specifically, the 360-degree direction on the image plane is divided into fixed angle intervals. For example, if each 45-degree interval is defined as a direction segment, it can be divided into eight direction categories, corresponding to: 0°~45°, 45°~90°, 90°~135°, 135°~180°, 180°~225°, 225°~270°, 270°~315°, and 315°~360°. The direction angle value of each differential path segment is classified into a specific direction category by determining its corresponding angle interval. For example, if a direction segment has an angle of 128 degrees, it is classified into the 90°~135° direction category. After classification, it is necessary to further determine the coordinates of each direction segment in the image. The spatial location of the marker is determined by dividing the image into regions based on its starting coordinates (x, y). The image spatial partition is divided into several grid units based on the image width and height. For example, at a resolution of 1920×1080, it is divided into nine regions with 640 pixels horizontally and 360 pixels vertically. The starting coordinates of each path segment are input and its region is determined. For example, if the starting coordinates are (850, 450), it belongs to the middle region. The region marker is combined with the path direction category number to establish the segment's belonging information in the image. Each path segment record includes its position number and direction category number in the image plane. Finally, the spatial location and direction information of all path segments are summarized to form an adjustment direction marker group, which is used to mark all regions and angle belonging relationships in the image that need to be adjusted in direction.

[0044] Please see Figure 5 The specific steps of S4 are as follows: S401: Based on the adjustment direction mark group, convert the direction marks into an angle step sequence and input it into the rotation adjustment platform. Acquire image frames at each angle step, and perform edge extraction and direction sequence reconstruction on the images to obtain the rotation step edge structure set. First, the angle information corresponding to each direction marker is extracted and converted into a standardized angle step sequence. During execution, the starting point and step interval of the corresponding rotation angle are determined based on the direction category and spatial location of the direction marker. For example, for a direction marker labeled "Central Region -45° Direction," the starting angle is set to 0°, with a step of 1° per step and a step range of ±10°. This generates a step sequence of 21 angle points: [-10°, -9°, ..., 0°, ..., +10°]. This step sequence is input to a controllable rotation adjustment platform. The platform pauses and maintains a stable position at each set angle for a set time, for example, a stabilization wait time of 1 second. Then, an image frame corresponding to the current angle is acquired using an image acquisition device. Each frame of the image is associated with and stored one by one with its corresponding angle. After the acquisition is completed, edge extraction processing is performed on each frame of the image. Using the pre-set edge extraction rules, the boundary area of ​​the workpiece is identified in the image, the coordinates of all edge points are extracted and the edge direction information is recorded. The edge points are sorted one by one to construct a direction sequence. This direction sequence is constructed based on the direction change value between adjacent points. For example, the direction sequence of the edge points of a certain segment in the image is 35°, 36°, 37°, and 39°. This is used to construct a direction sequence for subsequent analysis. The edge direction data of the image under all angle steps are indexed to form a rotation step edge structure set. Each set of data contains information such as angle value, image number, edge point set, and direction sequence, and is stored in angle order to form a set structure.

[0045] S402: Call the rotation step edge structure set, compare the boundary direction sequence in the image, calculate the rate of change of the angle between the direction vectors of adjacent frames, and locate the region with the slowest direction change based on the minimum rate threshold to obtain the image direction stable interval group. The image frame orientation sequence is read sequentially under adjacent angle steps. Angle calculation is performed on all corresponding orientation vectors in the orientation sequence between two frames. By comparing the angle difference between edge orientation vectors under adjacent angles, the magnitude of orientation change caused by the angle step is obtained. Then, all angle change values ​​are compared with the angle step difference to obtain the orientation change rate per unit angle. For example, if two frames have steps of 0° and 1° respectively, and the average angle change of the edge direction is 2.5°, then the rate is 2.5° / 1°. This operation is performed on all adjacent frame pairs to form a set of angle change rate sequences. Each value in this sequence represents the response intensity of the edge structure direction within a certain angle interval. Subsequently, the following settings are defined... The minimum rate threshold is used as a screening criterion. If the angle change rate of a certain frame pair is lower than the threshold, the image edge structure under that angle step is considered to be stable. The minimum rate threshold needs to be set empirically based on resolution and edge change sensitivity. For example, in actual experiments, it was found that the edge direction of regions with a rate lower than 0.8° / step is basically consistent. Therefore, the threshold is set to 0.8. All angle segments with a change rate lower than 0.8 are recorded as candidate regions with their corresponding angle ranges. Multiple consecutive frame pairs that meet the conditions are merged into a stable interval, and finally, a stable interval group of image direction is obtained. Each group of data includes the starting angle, ending angle, average rate value and corresponding image frame sequence number.

[0046] S403: Based on the stable interval group of the image direction, determine whether the image frame sequence in each region meets the angle step continuity condition, filter out all complete region segments of the sequence, extract the start and end range of the angle, and generate the adjustment direction segment; The continuity of the image frame sequence within each stable interval is determined. The angle markers of all frames within the interval are read to check if the stepping is continuous, i.e., whether the angle difference between the current angle and the previous frame is consistent. If an angle jump or discontinuity is found, for example, if the stepping should be 1° but jumps directly from 3° to 6°, the segment is determined not to meet the continuity condition and is not retained. Only the segment with continuous stepping without jumps is retained. Then, the starting angle and ending angle of each stable segment that meets the continuity condition are counted. For example, if a segment steps continuously from -4° to +3° and the angular rate is lower than a set threshold, then its angle start and end range is -4° to +3°. All segment that meets the condition is numbered and marked according to the angle interval, and its image frame index, corresponding angle range, and stable state identifier are recorded. Finally, all image segments that meet the angle stepping continuity and direction stability are sorted and output to form adjustment direction segments. Each segment corresponds to a suggested adjustment angle interval, which is convenient for subsequent angle correction or pose fine-tuning operations.

[0047] Please see Figure 6 The specific steps of S5 are as follows: S501: Based on the adjustment direction segment, divide the image into sub-blocks in the corresponding image frame, extract the edge direction on the sub-block splicing boundary, calculate the boundary direction curvature value, make a continuity judgment on the curvature direction change value between adjacent sub-blocks, and establish a splicing boundary curvature direction set. First, select the corresponding image frame and divide it into several image sub-blocks according to a unified image segmentation rule. The segmentation operation is carried out proportionally according to the image size. Usually, the size of each sub-block is set to a fixed number of pixels, such as 200×200 pixels, which is suitable for 1920×1080 resolution images and can generate approximately 9 columns × 5 rows, totaling 45 sub-block regions. After segmentation, process the stitching boundary of each pair of adjacent sub-blocks in turn, extract all pixels on the stitching boundary, and perform an edge direction extraction operation on each pixel. The direction information is the edge gradient direction angle. This direction value is obtained by reading the edge direction layer already established in the image. For example, if the direction of continuous points extracted on the boundary of a certain sub-block is 42°, 45°, 47°, 44°, 43°, use this sequence as input to perform a curvature calculation operation. The curvature is calculated based on the discrete change degree of the edge direction change trend. Specifically, the difference between each pair of adjacent angles in the direction angle sequence is taken and the absolute value is then divided. The stability of these differences is analyzed. For example, if the change in adjacent directions is always within ±3°, the curvature value is considered small. If the change between two points exceeds 10°, it indicates that the local curvature is large. The standard division of curvature values ​​is set according to the range of changes in the actual edges in the analyzed image. Generally, a curvature of 0° to 5° for each pair of directional angle differences is considered low curvature, 5° to 10° is considered medium curvature, and more than 10° is considered high curvature. The same calculation is performed on all boundaries, and the difference in the curvature values ​​of the boundary directions between adjacent sub-blocks is judged. If the average curvature difference between the boundaries of two spliced ​​sub-blocks is less than the set curvature continuity threshold, the connection is judged to be curvature continuous; otherwise, it is marked as curvature discontinuous. The curvature continuity threshold is determined empirically based on the difference distribution of typical continuous structural boundaries in the image, and is generally set to 5°. That is, when the curvature difference between boundaries is greater than 5°, it is marked as discontinuous. Finally, the curvature calculation results and continuity judgments of all sub-block boundaries are summarized to form a set of spliced ​​boundary curvature directions.

[0048] S502: Call the set of curvature directions of the splicing boundary, construct the curvature change path map in sequence in the splicing path of the image sub-blocks, calculate the curvature slope change value of adjacent boundary segments and compare it with the curvature change judgment threshold, identify the corresponding position of the change slope point, and obtain the curvature change point index set. A complete image sub-block stitching path is constructed according to the sub-block connection order. Within each path, the curvature data of adjacent boundary segments are arranged sequentially to construct a curvature change path diagram. The path diagram uses the image stitching order as the main line, with nodes representing sub-block boundaries and edges representing curvature jumps between adjacent boundaries. Then, the curvature slope change value is calculated for each pair of adjacent boundary segments in the path diagram. This is calculated by dividing the curvature increment from one boundary segment to the next by the path length covered by the two segments, obtaining the curvature change rate per unit length. A larger value indicates a more pronounced slope abrupt change. After calculating the curvature slope change values ​​for all path segments, the change values ​​need to be judged to identify locations with significant abrupt changes, and a curvature abrupt change judgment is set. A threshold is used as the criterion for judgment. The threshold setting needs to be determined in combination with the degree of natural boundary fluctuation in the image. Through statistical analysis of a large number of image samples, it was found that the rate of curvature change is mainly no more than 0.08 under normal circumstances. Therefore, the threshold is set to 0.1. That is, when the rate of curvature change of an adjacent boundary exceeds 0.1, it is considered that a sudden change has occurred. All path nodes that meet this condition are marked as abrupt slope points. The position information of these nodes is extracted, including the number of the boundary segment in the stitching path, the corresponding image sub-block number, the start and end pixel positions and curvature value changes. Finally, a curvature abrupt change point index set is formed. This set contains all abnormal change points and their positions in the image structure, which facilitates further localization and accuracy analysis.

[0049] S503: Based on the curvature change point index set, retrieve the corresponding splicing path number and perform path direction backtracking positioning, extract the index number of adjacent image sub-blocks at the change point and mark them, aggregate all marking results, and establish a meshing accuracy measurement and calibration scheme. The process involves sequentially reading the splicing path number corresponding to each mutation point and locating the mutation point's position within the splicing path. Then, using the mutation point as the center, the process traces back its adjacent connections, extracting the numbers of the two sub-blocks connected to the mutation point. This identifies the sub-block combination to which the point belongs. For example, if the mutation point is located at path number P8, and the corresponding sub-blocks are C3 and C4, then this point is marked as the C3-C4 splicing mutation pair. This process is repeated for all mutation points, recording and marking each pair of mutated sub-block combinations. Subsequently, an aggregation operation is performed on all marked sub-block pairs, merging all duplicate or adjacent regions and sorting them according to image coordinates to ensure spatial continuity and completeness in the subsequent output. Based on the aggregation results, an index table is established for each mutation sub-block pair. The table contains information such as sub-block number, splicing boundary position, mutation point direction difference, curvature change rate value, and image frame number. Based on the data in this table, a meshing accuracy measurement and calibration scheme is automatically generated. The calibration scheme clearly indicates the splicing boundary position of all sub-blocks involved in curvature mutations, forming a list of areas that need to be focused on detection or correction. At the same time, a directional mutation description is added to each marked point, which records the mutation value, the location of the occurrence area, the recommended calibration method, and the angle adjustment range. The scheme is finally output in a structured manner to guide image processing in performing directional calibration and error compensation operations on the workpiece meshing area.

[0050] 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 machine vision-based method for measuring and calibrating the meshing accuracy of a gear transmission system, characterized in that, Includes the following steps: S1: Extract contact edge points in the image, calculate the gradient direction of adjacent pixels, mark the angle direction, analyze the trend of direction change, remove abrupt points, construct direction segments and inter-frame indexes according to frame order, and generate edge direction segment sequence. S2: Based on the edge direction segment sequence, locate the intersection of the tooth profile, extract the longitudinal coordinates, analyze the changes in the path and distance of the intersection points within the frame, remove the sequences with no changes, evaluate the remaining fluctuation density, locate the tooth axis, and obtain the axial misalignment structure. S3: Based on the axial misalignment structure, extract the tooth tip profile curve, compare it with the standard path, calculate the point-to-point spatial difference, extract the direction and length of the difference segment, organize the regional structure and mark the discrete direction to form an adjustment direction mark group; S4: Based on the adjustment direction marker group, the direction markers are converted into a fine-tuning angle sequence, image frames are acquired and the direction changes are compared, stable segments are identified according to the minimum rate threshold, and adjustment direction segments are generated. S5: Based on the adjustment direction segment, extract the curvature direction of the image sub-block splicing boundary, analyze the direction continuity, identify the slope change point, trace back the path to locate the sub-block index, and establish a meshing accuracy measurement and calibration scheme.

2. The machine vision-based gear transmission system meshing accuracy measurement and calibration method according to claim 1, characterized in that, The edge direction segment sequence includes edge point direction angles, continuous direction segment numbers, inter-frame correspondence, and structural path indexes. The axial misalignment structure includes tooth intersection coordinates, frame sequence distribution trajectory, position fluctuation amplitude, and misalignment positioning segments. The adjustment direction marker group includes boundary difference segments, direction change, length coverage, and discrete direction angle identifiers. The adjustment direction segment includes angle step sequence, direction change rate, stable direction segment, and continuity verification results. The meshing accuracy measurement and calibration scheme includes sub-block splicing numbers, curvature change paths, abrupt connection point indexes, and interface marking information.

3. The machine vision-based gear transmission system meshing accuracy measurement and calibration method according to claim 1, characterized in that, The unchanged sequence refers to a continuous data sequence in which the path and distance of the intersection points within a frame remain constant.

4. The machine vision-based gear transmission system meshing accuracy measurement and calibration method according to claim 1, characterized in that, The minimum rate threshold refers to the minimum rate threshold used to determine the stability of directional changes.

Images