Electric tricycle steel ring automatic seam detection method based on image processing

Abstract

The application relates to the technical field of image recognition, and provides an automatic joint detection method for a steel ring of an electric tricycle based on image processing. The method comprises the following steps: image acquisition is performed on the surface of the steel ring of the electric tricycle to obtain a steel ring surface image sequence; edge analysis is performed on the steel ring surface image sequence to generate a joint feature atlas; morphological analysis is performed based on the joint feature atlas to identify a plurality of joint parameters; quality determination is performed according to the plurality of joint parameters to generate a joint quality report; a joint searching signal is triggered according to the joint quality report; and automatic joint detection is performed on the steel ring of the electric tricycle based on the joint searching signal. The application solves the technical problem that the traditional image recognition technology is interfered by complex textures on the surface of the steel ring, joint feature extraction is blurred, and low detection efficiency and high missing detection rate are caused, and achieves the technical effect that through dynamic edge analysis driven by image recognition combined with multi-scale morphological processing, the joint detection precision and the automation level of the steel ring are improved, and the missing detection rate is reduced.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of image recognition, in particular to an automatic joint detection method for a steel rim of an electric tricycle based on image processing. BACKGROUND

[0002] As a core structure for bearing the load of a vehicle, the joint quality of the steel rim directly affects the overall mechanical strength and safety. Traditional detection processes rely on manual visual inspection or contact-type measurement tools, which have the problems of low efficiency, strong subjectivity, and easy omission of small defects. Especially in large-scale production, the surface of the steel rim often has complex textures due to casting residues, oxidation stains, and light interference, which makes it difficult for conventional image recognition algorithms to accurately extract joint features, and often causes edge blurring and false judgment. Existing automated detection equipment mostly uses fixed two-dimensional imaging schemes, which cannot adapt to the curved surface characteristics of the steel rim, and lack the ability to quantitatively analyze multi-scale joint parameters (such as straightness, width deviation, and angle deviation), making it difficult to meet high-precision industrial quality inspection standards. In addition, the manual re-inspection process increases the downtime of the production line, severely restricting the improvement of production efficiency. Therefore, there is an urgent need for a technology that can automatically, efficiently and accurately detect the joint of the steel rim to improve the automation level of the production line, reduce the omission rate, and ensure the quality of the product. SUMMARY

[0003] The application provides an automatic joint detection method for a steel rim of an electric tricycle based on image processing, which aims to solve the technical problems of low detection efficiency and high omission rate caused by the extraction of blurred joint features due to the complex texture interference on the surface of the steel rim in traditional image recognition technology.

[0004] The application provides an automatic joint detection method for a steel rim of an electric tricycle based on image processing, which includes: image acquisition of the surface of the steel rim of the electric tricycle to obtain a sequence of steel rim surface images; edge analysis of the sequence of steel rim surface images to generate a joint feature map, morphological analysis based on the joint feature map to identify multiple joint parameters; quality determination based on the multiple joint parameters to generate a joint quality report, triggering a joint detection signal according to the joint quality report, and automatic joint detection of the steel rim of the electric tricycle based on the joint detection signal.

[0005] One or more technical solutions provided in the application have at least the following technical effects or advantages:

[0006] The above-mentioned image processing-based automatic joint detection method for electric tricycle steel rings first captures the surface of the electric tricycle steel ring through an image acquisition device to obtain a series of steel ring surface images. Then, edge detection is performed on these images to generate a joint feature map. Based on the map, morphological analysis is performed to identify a plurality of joint parameters. Then, the joint quality is evaluated according to these parameters, and a quality report is generated. Finally, a joint detection signal is triggered according to the quality report, and the steel ring is automatically detected through the signal, thereby completing an efficient and accurate joint detection process.

[0007] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the content of the specification can be implemented, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0009] Figure 1 A flowchart of an embodiment of an image processing-based automatic joint detection method for electric tricycle steel rings.

[0010] Figure 2 A flowchart of an embodiment of an image processing-based automatic joint detection method for electric tricycle steel rings. DETAILED DESCRIPTION

[0011] The embodiments of the present application provide an image processing-based automatic joint detection method for electric tricycle steel rings, which solves the technical problems of low detection efficiency and high missed detection rate caused by blurred joint feature extraction due to complex texture interference on the surface of the steel ring in traditional image recognition technology.

[0012] The technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0013] It is to be understood that the terms "including", "comprising", "having" and

[0014] As shown in the embodiments, the present application provides an image processing-based automatic seam detection method for electric tricycle steel rims, which comprises the following steps: Figure 1 As shown in the embodiments, the present application provides an image processing-based automatic seam detection method for electric tricycle steel rims, which comprises the following steps:

[0015] Image acquisition is performed on the surface of the electric tricycle steel rim to obtain a sequence of steel rim surface images.

[0016] In the embodiments of the present application, in order to avoid the influence of factors such as shadow or reflection on the image quality of the electric tricycle steel rim, a ring array light source is used to perform uniform illumination on the surface of the electric tricycle steel rim. The ring array light source can provide omnidirectional illumination to ensure uniform distribution of light, and the collected images are clear and distortion-free. After uniform illumination, an image acquisition device (such as an industrial camera) is set to take pictures of the surface of the electric tricycle steel rim to obtain a series of continuous steel rim surface images. These images can record all the details of the steel rim surface and form an image sequence, i.e., a sequence of steel rim surface images, which provides necessary raw data for subsequent image processing and seam detection.

[0017] Further, as shown in the embodiments, the present application provides an image processing-based automatic seam detection method for electric tricycle steel rims, which comprises the following steps: Figure 2

[0018] The electric tricycle steel rim is fixed on the center axis of the rotating platform, the ring array light source is used to irradiate the surface of the steel rim, and the irradiation light source parameters are determined, which include the light source stroboscopic frequency data; the rotating platform is started to rotate the electric tricycle steel rim based on the irradiation light source parameters to obtain the rotating speed data; the scanning mode of the line array industrial camera is triggered, the electric tricycle steel rim is scanned through the scanning mode to obtain a set of steel rim scanning parameters; the set of steel rim scanning parameters is adjusted according to the light source stroboscopic frequency data and the rotating speed data to generate a set of steel rim full-circle surface images; and the set of steel rim full-circle surface images is integrated according to the scanning time sequence to generate the sequence of steel rim surface images.

[0019] ​Preferably, first, the electric tricycle rim is fixed on the central shaft of the rotating platform through the fixing device to ensure that the rim remains stable and does not displace during rotation, then the rim surface is uniformly irradiated using the annular array light source, and the irradiation parameters of the light source are determined, wherein the position and incident angle of the annular light source are set to preset values to ensure that the light is uniformly irradiated on the surface of the rim; the light source irradiation parameters include a stroboscopic frequency data, which is matched with the rotation speed to ensure that the stroboscopic light source can be synchronized with the angular velocity of the rotating platform at each frame of image acquisition. Subsequently, the rotating platform is driven to rotate by using a servo motor, and the angular displacement of the rim is fed back in real time as rotation by using an absolute value encoder, and the motor control system adjusts the angular velocity of the rotating platform in real time according to the feedback information to ensure that the rim rotates at a constant angular velocity, and the real-time angular velocity is stored as rotation speed data. In the rotation starting stage, acceleration ramp control is used to smoothly accelerate to avoid image distortion caused by instantaneous acceleration, and the number of rotations is set to an integer multiple to ensure that the starting position of the rim is aligned with the ending position to avoid angle errors in the overlapping area of the image. After the rotating platform is started, the line array industrial camera triggers the scanning mode to gradually scan the surface of the rim. In the scanning process, the line array camera calculates the real-time scanning frequency according to the diameter of the rim (the product of the angular velocity and the diameter is divided by twice the maximum scanning line spacing) to ensure that the spacing between adjacent scanning lines is not more than 0.05 mm to obtain high-resolution image data, and at the same time, the parameters in the scanning process and the collected image data are added to a set to obtain a rim scanning parameter set, such as scanning position data and image data. The camera will immediately enter the next row of scanning every time a row of data is scanned to ensure the continuity and accuracy of image acquisition. Subsequently, the image data in the rim scanning parameter set is analyzed to determine whether there are blurred images. If there are, it means that the current image acquisition has motion blur due to the rotation of the rim. At this time, the rotation speed data is used to control the rotation speed of the rim. If the image is blurred due to too fast rotation, the speed can be adjusted to eliminate the motion blur. At the same time, the rotation speed is used to improve the stroboscopic frequency of the light source (the linear velocity is divided by the camera pixel spacing), and the image acquisition time is synchronized with the stroboscopic frequency of the light source to ensure that the brightness of each frame of image remains uniform, avoid uneven image brightness caused by inconsistent stroboscopic, and thus maintain the uniformity of image grayscale. After adjustment, image acquisition is performed again to obtain images which will serve as a rim full-surface image set. Finally, the images in the rim full-surface image set are arranged in chronological order to form a continuous rim surface image sequence to provide accurate data support for subsequent joint detection and analysis.

[0020] Edge analysis is performed on the rim surface image sequence to generate a joint feature map, and morphological analysis is performed based on the joint feature map to identify a plurality of joint parameters.

[0021] In one embodiment, after obtaining the steel ring full circumference surface image sequence, first, edge analysis is performed on the image sequence, and the edge profile of the steel ring surface joint is identified by applying an edge detection algorithm (such as Canny edge detection), which represents the position and shape of the joint, and the edge data of each frame in the image sequence will be used as the basis for subsequent feature extraction. Subsequently, the joint edge profile data extracted from the image is converted to a polar coordinate system, and a plurality of joint profile data such as radial distance data and angular position data is determined, and then by merging edge data at different angles, the features of all joints on the steel ring surface are spliced into a complete joint feature map, which shows the joint feature information of the entire steel ring. Then, based on the joint feature map, morphological analysis is performed on the joints in the map, in which the edge profile data in the joint feature map is enhanced, and the enhanced data is used to calculate the straightness evaluation index, joint width data and angle deviation value, which will jointly form a plurality of joint parameters as an important basis for subsequent quality evaluation, and provide a reference for subsequent automatic joint detection and joint quality control.

[0022] Further, the application provides an edge analysis method for the steel ring surface image sequence to generate a joint feature map, which comprises:

[0023] The steel ring surface image sequence is analyzed for image overlap according to adjacent image frames to determine a plurality of overlapping regions; the steel ring surface image sequence is registered according to the plurality of overlapping regions to construct a steel ring surface development map; edge detection is performed according to the steel ring surface development map to extract joint edge profile data; a polar coordinate system is constructed based on the steel ring surface development map, and the joint edge profile data is converted to the polar coordinate system to determine a plurality of joint profile data, which includes radial distance data and angular position data; and the joint feature is spliced according to the radial distance data and the angular position data to generate the joint feature map.

[0024] Preferably, firstly, the adjacent image frames in the steel ring surface image sequence are subjected to overlap analysis, the relative displacement of each image frame is compared, and the overlap area between them is determined. The overlap area refers to the part shared by adjacent images. These areas provide a reference for subsequent registration and stitching, and ensure the alignment between images. Subsequently, the steel ring surface image sequence is registered according to the determined multiple overlap areas. This step eliminates the deviation caused by the difference in shooting angles or the rotation of the steel ring by aligning the overlapping parts of adjacent image frames. After registration, all images are stitched into a unified steel ring surface development map, which shows the joint features of the entire steel ring surface. Then, based on the constructed steel ring surface development map, an edge detection algorithm (such as Canny edge detection) is used to extract the edge profile data of the joint. Edge detection can identify the clear profile of the joint, especially in the joint area of the image, generating a set of joint edge profile data representing the position and shape of the joint. Then, a polar coordinate system is established based on the center of rotation of the steel ring. In the polar coordinate system, each edge point of the joint is represented by a radial distance r and an angle θ. The center of rotation of the steel ring is located at the origin of the polar coordinate system. The shape of the joint is defined by the distance from the origin to the joint edge. Then, from the steel ring surface development map, the edge profile data of the joint is extracted. Assuming that the coordinates of the joint edge point in the Cartesian coordinate system are (x, y), where x is the horizontal coordinate and y is the vertical coordinate. For each joint edge point, its radial distance is calculated according to its coordinates (x, y) in the Cartesian coordinate system by the formula and its angle is calculated according to the formula To ensure the continuity and accuracy of the edge data in the polar coordinate system, when selecting the joint edge point, it is resampled according to a predetermined resolution (such as an angle resolution of 0.1°). That is, every 0.1° in the circumferential direction, a joint edge point is selected, and the angle and radial distance of the point are generated. Through resampling, the angle and corresponding radial distance of each joint edge point are obtained, which will be used as multiple joint profile data to participate in the subsequent atlas construction process. Then, the image is stitched according to the radial distance data and angular position data of the joint. In the image stitching process, angle jump errors may occur in the overlapping areas. At this time, through a data fusion algorithm (such as weighted average method, B-Spline interpolation method, etc.), the overlapping areas are smoothed to ensure the continuity and error-free nature of the joint profile in the polar coordinate system after stitching. After stitching, an angle-radial distance curve is obtained, which is the joint feature atlas. The horizontal axis of the joint feature atlas represents the angle θ (i.e., the position on the steel ring), and the vertical axis represents the radial distance r (i.e., the distance from the center of the steel ring to the edge of the joint). The position of each point on the atlas is the radial distance and angle information of a joint edge point.

[0025] Through the above steps, the complete information of the bead surface joint can be effectively extracted and spliced, ensuring the accuracy and continuity of the joint features, and providing accurate basis for subsequent quality detection and automatic joint searching.

[0026] Further, the present application provides morphological analysis based on the joint feature map, and identifies a plurality of joint parameters, the method comprising:

[0027] Based on the joint edge contour data, multi-scale edge enhancement processing is performed to extract edge contour pixel data of a plurality of joint regions; the radial distance data and the angular position data are analyzed according to the edge contour pixel data by traversing the joint feature map, and a data mapping relationship network is constructed; the joint feature map is analyzed according to the data mapping relationship network, and the plurality of joint parameters are identified.

[0028] Optionally, when performing morphological analysis based on the joint feature map, first, multi-scale edge enhancement processing is performed on the image based on the joint edge contour data to improve the edge definition of the joint region. In this process, edge response features are extracted by non-subsampled shearlet transform (NSST) to enhance the contrast of the joint edge, and polygon data of the joint contour is generated by using edge tracking technology to more accurately describe the shape and position of the joint. Based on these polygon data, complete edge contour pixel data of a plurality of joint regions are generated through steps such as amplitude calculation, retrieval screening, and pixel-level connection to provide a basis for subsequent joint feature analysis. Subsequently, joint feature points in the joint feature map are gradually obtained, as well as corresponding radial distance data and angular position data, which will be used for subsequent mapping analysis to form the spatial distribution of the joint. The pixel position of the joint edge point and the pixel position in the edge contour pixel data are used to correspond the radial distance data, angular position data, and geometric features of the same joint edge point, and a data mapping relationship network is constructed. Through this mapping relationship, the local features of the joint can be combined with the global features to realize global analysis of the joint morphology. After the data mapping relationship network is constructed, morphological analysis of the joint feature map is performed, including analysis of the angle deviation value, analysis of the straightness evaluation index, and analysis of the joint width data. Then, these analysis results are arranged to obtain a plurality of joint parameters to comprehensively evaluate the geometric features of the joint and ensure that the joint quality of the bead meets the requirements.

[0029] Further, the present application provides multi-scale edge enhancement processing based on the joint edge contour data to extract edge contour pixel data of a plurality of joint regions, the method comprising:

[0030] A non-subsampled wavelet transform is performed on the surface unfolded map of the steel ring to extract edge response features; edge tracking is performed based on the edge response features to generate seam contour polygon data; the gradient magnitude of the edge points is calculated based on the seam contour polygon data, and the seam edge contour data is searched and filtered according to the gradient magnitude of the edge points to determine multiple valid edge points; the multiple valid edge points are connected at the pixel level to obtain the edge contour pixel data of multiple seam regions.

[0031] Optionally, non-subsampled wavelet transform (NSST) is applied to the surface unfolded image of the steel ring. NSST is a multi-scale transform method that effectively extracts edge features at different scales from the image. Unlike traditional wavelet transform, NSST does not perform subsampling, thus preserving high-frequency information and better retaining edge details. In this process, by selecting appropriate wavelet basis functions, the surface unfolded image of the steel ring is transformed, generating multi-level transform coefficients. These coefficients contain the edge response features of the image, corresponding to changes in the seam edges and reflecting the structural characteristics of the seam at different scales. By analyzing these features, the basic contour of the seam region in the image can be understood. Subsequently, based on the extracted edge response features, the seam edges are tracked. Common edge tracking methods include gradient methods, direction methods, or threshold-based edge extraction algorithms. These methods can find the accurate boundary of the seam through the gradient information of the image and perform precise contour tracking of the seam. After edge tracking is complete, the seam edge data is converted into polygon data. This polygon data consists of continuous points along the seam edges, with each polygon vertex representing an edge point of the seam. These points are arranged sequentially to form the overall contour of the seam. Then, for each polygon data point of the seam contour, the gradient magnitude of each edge point is calculated. Specifically, the formula is first used... Calculate the gradient magnitude of the image at that point. ,in, and These are the brightness gradients of the image in the x and y directions, respectively. Then, adding 90° to the tangent direction of the seam edge gives the normal direction of the seam edge. Then use the formula Calculate the gradient magnitude in the normal direction. ,in, , where is the unit vector in the normal direction. Then, based on the gradient magnitude of the normal direction at the edge points, the seam edge points are filtered. Points with larger gradient magnitudes and significant brightness changes—that is, points whose normal direction gradient magnitude is greater than or equal to a gradient magnitude threshold—are considered valid edge points. These points accurately reflect the position and shape of the seam. Finally, multiple selected valid edge points are connected pixel-level to form a continuous edge contour. Through this connection, complete edge contour pixel data for multiple seam areas can be obtained. This data provides the necessary foundation for subsequent seam quality analysis, defect detection, and repair, ensuring the accuracy of image recognition, improving the precision of steel ring seam detection, and reducing the false negative rate.

[0032] Furthermore, this application provides a method for analyzing the seam morphology of the seam feature map according to the data mapping relationship network and identifying the multiple seam parameters, including:

[0033] Transformation detection is performed on the edge contour pixel data based on the polar coordinate system to determine the seam centerline. Deviation analysis is then performed based on the seam centerline to generate an angle deviation value. A sliding window is set according to the data mapping relationship network, and the edge contour pixel data is synchronized to the sliding window for calculation to construct a straightness evaluation index. The average distance between the two edges of the seam is retrieved and calibrated to determine the seam width data. The straightness evaluation index and the seam width data are corrected according to the angle deviation value to obtain the multiple seam parameters.

[0034] Optionally, based on the seam edge contour data in polar coordinates, the centerline of the seam is determined. The centerline is typically located along the straightest part of the seam. Through linear fitting, the position of the centerline is determined, and the offset of each point on the seam relative to the centerline is calculated, generating an angular deviation value. This deviation value measures the degree of eccentricity of the seam during rotation, reflecting its straightness and geometric accuracy. Subsequently, a sliding window is set in the data mapping network to analyze the local features of the seam. The window size can be adjusted according to actual needs to ensure coverage of different areas of the seam. Within the sliding window, a straightness evaluation index is constructed by calculating the standard deviation of the radial distances of the edge points. The standard deviation reflects the degree of fluctuation of the seam within the window; a relatively straight seam has a small standard deviation, while a large standard deviation indicates significant curvature or offset. This index quantifies the straightness of the seam and assesses its deviation. Next, based on the seam edge contour data, the distance between the two edge points at different locations of the seam (the difference in radial distance between the two edge points of the seam) is obtained. The average spacing between the two edges of the seam is calculated, and this average spacing is calibrated as the seam width data. The seam width data can help further judge the quality of the seam, especially when the seam width exceeds the standard or is uneven; the calibrated data can serve as a basis for anomaly detection. In actual production, due to installation eccentricity, the width and straightness of the seam may deviate. Therefore, the calculated angular deviation value is used to correct the straightness evaluation index and the seam width data. For the straightness evaluation index, an angular deviation value is added to the angle of each edge point, and the radial distance of each edge point is recalculated. Then, the standard deviation is calculated based on the corrected radial distance to obtain the corrected straightness evaluation index. For the seam width data, the new distance between the two edge points is calculated using the corrected radial distance, and then the average of these distances is calculated to obtain the corrected seam width data. Finally, multiple seam parameters are generated based on the corrected seam width data and the straightness evaluation index. In summary, using a sliding window to calculate straightness evaluation indicators, combined with joint width, allows for a comprehensive assessment of joint quality, ensuring it meets design and production requirements. By correcting for errors caused by installation eccentricity, the accuracy of joint quality assessment is further improved, providing reliable data support for subsequent quality control and automated inspection.

[0035] The quality is determined based on the multiple joint parameters, a joint quality report is generated, a joint detection signal is triggered according to the joint quality report, and automatic joint detection is performed on the steel rim of the electric tricycle based on the joint detection signal.

[0036] In one embodiment, after the seam parameters are calculated, a comprehensive quality score is calculated by judging the seam quality of multiple seam parameters and then compared with a predetermined quality standard to determine whether the seam meets production requirements. If the seam parameters exceed the quality standard, it is judged as unqualified; otherwise, the seam is considered to meet the standard. Based on the quality judgment results, a detailed seam quality report is generated, which records multiple defect locations to provide guidance for subsequent defect handling. After obtaining the seam quality report, a seam-finding signal is triggered based on the results. If a problem exists in the seam, this signal will be used as the trigger condition to start the automatic seam-finding detection process of the electric tricycle steel rim, accurately re-inspecting or correcting the seam area to ensure that the seams of each steel rim meet the quality requirements during the production process. This process realizes automated detection and control of seam quality, improving production efficiency and product quality consistency.

[0037] Furthermore, this application provides a method for determining quality based on the multiple joint parameters and generating a joint quality report, including:

[0038] A joint quality evaluation matrix is ​​constructed. Based on the joint quality evaluation matrix, a weighted analysis is performed on the multiple joint parameters to obtain multiple joint weight coefficients. The multiple joint parameters are then fused and analyzed according to the multiple joint weight coefficients to determine a comprehensive quality score. Based on the comprehensive quality score, the joint areas are traversed for judgment, and multiple substandard joint areas are extracted. The multiple substandard joint areas contain multiple joint defect coordinates. Spatial clustering analysis is performed on the multiple substandard joint areas to obtain isolated defect classes and continuous defect classes. The multiple joint defect coordinates are mapped to the polar coordinate system according to the isolated defect classes and the continuous defect classes to generate multiple defect location markers. The multiple defect location markers are added to the joint quality report.

[0039] Optionally, a joint quality evaluation matrix is ​​constructed. This matrix includes quality standards for multiple joint parameters, such as width tolerance threshold, maximum permissible deviation of straightness, and continuous defect length limit. The width tolerance threshold sets the allowable deviation range for the joint width, used to determine whether the joint meets the predetermined standard width. If the width exceeds this tolerance range, it is considered unqualified. The maximum permissible deviation of straightness sets the maximum permissible straightness deviation of the joint. Joints exceeding this deviation range are considered curved or uneven, and therefore unqualified. The continuous defect length limit defines the maximum length limit for continuous defects on the joint. If continuous defects on the joint exceed this length, it is considered unqualified. Based on the above joint quality evaluation matrix, a weighted analysis is performed on multiple joint parameters. Each joint parameter, such as joint width, straightness, and defect length (continuous areas where the joint width exceeds the tolerance threshold or the straightness deviation exceeds the maximum permissible deviation), is assigned a weight coefficient according to its importance in the overall quality judgment. The weight coefficient can be set according to actual needs and production requirements to reflect the relative impact of each parameter on quality. Subsequently, based on the weighted joint parameters, the various joint parameters are fused and analyzed to generate a comprehensive quality score. The comprehensive quality score is the weighted sum of all joint parameters, reflecting the overall quality of the joints. It is important to note that before weighting, the joint parameters are processed using the min-max normalization method to ensure that the data are on the same scale. Next, each joint region is traversed, and the comprehensive quality score of each joint region is compared with a preset acceptable threshold. Multiple joint regions below the threshold are selected as multiple non-compliant joint regions. These non-compliant joint regions contain the joint defect coordinates, which reflect the specific location of the joint and help locate problem areas. Then, K-means or DBSCAN is used to perform spatial clustering analysis on the extracted multiple non-compliant joint areas. The distance between each pair of non-compliant joint areas is calculated using Euclidean distance and compared with a preset connection judgment threshold. This classifies the defect points into isolated defect classes and continuous defect classes based on their spatial distribution. Isolated defect classes represent geographically dispersed and isolated defect areas, i.e., non-compliant joint areas that do not belong to any cluster. Continuous defect classes represent multiple joint defects that are spatially continuous, i.e., non-compliant joint areas that are clustered together. After obtaining the isolated and continuous defect classes, the coordinates of the joint defects in the isolated and continuous defect classes are mapped to a polar coordinate system, and a marker is generated for each defect location, recording the specific location and type (isolated defect or continuous defect), thus obtaining multiple defect location markers. Finally, the generated multiple defect location markers are added to the joint quality report, which will contain detailed information on all non-compliant joint areas, including the location, type, and corresponding quality score of each defect. This provides a detailed basis for subsequent quality repair, thereby improving the efficiency of quality control in the production process and ensuring that the joints of each steel ring meet quality standards.

[0040] Furthermore, this application provides a method for automatically detecting seam gaps in the steel rim of an electric tricycle based on a seam quality report-triggered seam detection signal, including:

[0041] When there are N defect location markers in the steel rim of an electric tricycle, the N defect location markers are discretely analyzed to generate discretely distributed defect data. Based on the discretely distributed defect data, a composite defect early warning signal is triggered. The composite defect early warning signal includes a seam-finding signal. The defects are classified according to the composite defect early warning signal to construct a defect severity level. The defect severity level is used as an index to match the sorting strategy database of the electric tricycle steel rim to generate a target sorting strategy. Based on the seam-finding signal, the target sorting strategy is executed to re-inspect and verify the defect seams. Based on the verification results, the electric tricycle steel rim is automatically checked and corrected using seam-finding detection.

[0042] Optionally, during the joint inspection of the electric tricycle's steel rim, defect locations in the joint area are identified and marked. When N defect location markers are detected, where N represents the number of discretely distributed defects on the steel rim (a positive integer greater than or equal to 3), discrete analysis is performed on the detected N defect location markers. The purpose of discrete analysis is to calculate the spatial distance between defect points through spatial analysis, generating discretely distributed defect data. This data includes the spatial location of the defects, the distance between each defect location and other defect locations, and the comprehensive quality score corresponding to each defect location. Subsequently, based on the discretely distributed defect data, a composite defect warning signal is triggered. This warning signal not only reflects the location of multiple defects but also indicates whether further automatic joint detection should be initiated through an internal joint-finding signal, ensuring that joint defects are effectively handled. When automatic joint detection is triggered, defects are classified according to the composite defect warning signal. The standard for defect classification is usually based on the number and distribution of joint defects. For example, if multiple defects are located in the same area, they may be considered serious defects. By analyzing the distribution of discrete defects, corresponding defect severity levels are constructed according to preset standards, such as minor, moderate, and severe defects. Then, based on the defect severity level, the most suitable sorting strategy is selected from the sorting strategy database through an index, serving as the target sorting strategy. This target strategy guides subsequent inspection and automatic seam correction. Next, based on the triggered seam correction signal and the target sorting strategy, the defective seams are re-inspected and verified by comparing manual inspection results with system inspection results, thereby generating verification results. This ensures that each seam meets quality standards, improving production efficiency and the consistency of steel ring product quality.

[0043] Furthermore, this application provides a method for matching the defect severity level as an index against a database of sorting strategies for electric tricycle steel rims to generate a target sorting strategy, the method comprising:

[0044] Based on the severity level of the defects, the sorting strategy database is matched to construct a rework priority level; according to the rework priority level, the coordinates of the multiple seam defects are synchronized to the automatic polishing workstation for recording, and a defect polishing file is generated; based on the defect polishing file, periodic analysis is performed to obtain the defect distribution pattern, and the defect distribution pattern is traced back to the sorting strategy database to select and determine the target sorting strategy.

[0045] Optionally, the severity level (minor, moderate, severe) of the seam defect is matched with the rework priority level rules recorded in the sorting strategy database to construct a rework priority level for each seam defect. The rework priority level reflects the urgency and repair priority of each defect. Severe defects (e.g., large or consecutive defects) will be assigned a higher rework priority, while minor defects (e.g., isolated small defects) will be assigned a lower rework priority. The rework priority level can be dynamically adjusted according to different production needs and standards to ensure that the defects that need the most repair are handled in a timely manner. Subsequently, according to the rework priority level, the coordinates of all seam defects requiring repair are synchronized to the automatic grinding workstation. These coordinates indicate the location of the defects. The workstation locates the defect area using precise coordinates and automatically starts the grinding operation to repair the defective area on the steel ring surface. During the grinding process, the automatic grinding workstation records all operation data in real time, including grinding time, tools used, grinding area, defect repair status, etc. All this data will generate a defect grinding file to track the repair history of defects and provide a basis for subsequent quality control. Next, a periodic analysis is performed based on the grinding records to assess the distribution of defects across different time periods and steel ring batches. Periodic analysis helps identify the recurrence of defects or potential problems in certain production stages. For example, if defects occur frequently within a certain period, it may be due to equipment failure or improper operation. Based on the results of the periodic analysis, the distribution patterns of seam defects are extracted. These patterns reveal the frequency, distribution area, and related factors (such as operation, equipment, and environment) of defects, providing data support for further production optimization. Once the defect distribution patterns are obtained, they are traced back to the sorting strategy database to find the most suitable sorting strategy related to defect type, distribution pattern, and periodicity. Cosine similarity can be used for matching. Based on the results of the backtracking analysis, the target sorting strategy that best matches the current defect distribution pattern is selected. This target sorting strategy includes changing the production process, adjusting equipment settings, adding quality checkpoints, and optimizing the rework process. This strategy will be applied to subsequent production stages to ensure more effective control measures are taken in areas with high defect rates, reducing subsequent rework and quality problems.

[0046] Furthermore, this application provides a method for re-inspecting and verifying defective seams by executing the target sorting strategy based on the seam-finding signal, and for automatically detecting and correcting seam-finding on the steel rim of an electric tricycle based on the verification results. The method includes:

[0047] Based on the seam-finding signal, reference markers are set in the non-seamless area of ​​the electric tricycle's steel rim. The repeatability of the rotating platform is calculated based on the reference markers. The multi-cycle target sorting strategy is executed to obtain multi-cycle sorting data. The multi-cycle sorting data is analyzed according to the repeatability of the rotating platform. The defective seams are evaluated based on the analysis results, and system detection results are generated. The manual detection results are compared with the system detection results for re-inspection and verification, and the verification results are generated.

[0048] Optionally, in the non-seamless areas of the electric tricycle's steel rim, reference markers are automatically set in these areas based on seam-finding signals. These reference points do not involve the seam area but are used to monitor the accuracy and stability of the steel rim during rotation. The reference markers are evenly distributed on the steel rim to ensure effective evaluation of the platform's positioning accuracy during rotation. By detecting the positional changes of the reference markers during multiple rotations, the repeatability of the rotating platform is calculated. Repeatability refers to the displacement fluctuation range of the same reference marker in different cycles. If the displacement of the marker is small, it indicates that the rotating platform has high accuracy; if the displacement is large, it indicates that the platform has repeatability errors, which may affect subsequent defect detection. Subsequently, based on the seam-finding signals, a target sorting strategy is executed. The target sorting strategy determines the processing method of the steel rim based on the quality of the seam and the type of defect, ensuring that each steel rim's defects are appropriately sorted and processed. During the execution of the target sorting strategy, sorting data from multiple cycles is obtained. The sorting data from each cycle records the steel rim's quality assessment and defect sorting information, including the location of seam defects and the number of defects sorted. Next, sorting data from multiple cycles is analyzed to calculate the standard deviation of positioning deviation and the number of defective items sorted across these cycles. These standard deviations reflect the fluctuations in sorting results across multiple cycles. Based on the calculated standard deviations, the quality of defective seams and the sorting effectiveness are evaluated. If the standard deviations are all within the tolerable range, it indicates consistent defect assessment and good seam quality. If the standard deviations are large, further evaluation of the seam quality is needed to determine if errors are caused by equipment instability or sorting strategy issues. System inspection results are generated based on the evaluation results. These results include a comprehensive evaluation of seam defects and defect location markings. Then, to verify the accuracy of the system inspection, these results are compared with manual inspection results. Manual inspection is typically performed by experienced technicians and serves as a reference standard. The system inspection results are compared with the manual inspection results point-by-point to ensure the reliability of automated inspection. During the comparison process, two metrics are calculated: the false negative rate and the false positive rate. The false negative rate refers to the proportion of undetected defects out of all defects; a high false negative rate indicates that some defects may be missed or not detected. The false positive rate refers to the proportion of falsely detected defects out of all defects; a high false positive rate means that some areas may be incorrectly marked as defects. By calculating these rates, the detection process can be further optimized to ensure the accuracy of the results. Finally, based on the calculations of the false negative and false positive rates, re-inspection verification results are generated. These results help evaluate the accuracy and stability of the detection, and are used to optimize automated detection strategies. If the false negative or false positive rate is high, the sorting strategy, the triggering method of the seam-finding signal, and the accuracy of the rotating platform will be adjusted according to the verification results to improve the accuracy of subsequent inspections, thereby ensuring the efficiency and accuracy of seam quality control.

[0049] In summary, the embodiments of this application have at least the following technical effects:

[0050] This application first acquires images of the surface of the steel rim of an electric tricycle, obtaining a sequence of steel rim surface images. Then, edge analysis is performed on the steel rim surface image sequence to generate a seam feature map. Morphological analysis is then performed based on the seam feature map to identify multiple seam parameters. Finally, a quality judgment is made based on the multiple seam parameters, generating a seam quality report. A seam-finding signal is triggered according to the seam quality report, and automatic seam detection of the electric tricycle steel rim is performed based on the seam-finding signal. These technical effects collectively solve the technical problem of low detection efficiency and high false negative rate caused by the complex texture interference of the steel rim surface in traditional image recognition technology, which leads to blurred seam feature extraction. This achieves the technical effect of improving the accuracy and automation level of steel rim seam detection and reducing the false negative rate through dynamic edge analysis driven by image recognition combined with multi-scale morphological processing.

[0051] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0052] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0053] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. An automatic seam detection method for steel rims of electric tricycles based on image processing, characterized in that, The method includes: Images of the steel rim surface of an electric tricycle were acquired to obtain a sequence of images of the steel rim surface; Edge analysis is performed on the image sequence of the steel ring surface to generate a seam feature map. Morphological analysis is then performed based on the seam feature map to identify multiple seam parameters. Based on the multiple seam parameters, a quality judgment is made, a seam quality report is generated, a seam detection signal is triggered according to the seam quality report, and automatic seam detection is performed on the steel rim of the electric tricycle based on the seam detection signal. The sequence of images obtained from the steel ring surface includes: The steel rim of the electric tricycle is fixed to the central axis of the rotating platform. The surface of the steel rim is illuminated by a ring array light source to determine the parameters of the illumination light source, which include the flicker frequency data of the light source. Based on the parameters of the illumination source, the rotating platform is started to rotate the steel wheel of the electric tricycle to obtain rotation speed data; The scanning mode of the linear industrial camera is triggered, and the steel rim of the electric tricycle is scanned through the scanning mode to obtain the steel rim scanning parameter set; The scanning parameter set of the steel ring is adjusted according to the light source flicker frequency data and the rotation speed data to generate a full circumference surface image set of the steel ring. The full circumference surface image set of the steel ring is integrated according to the scanning time sequence to generate the surface image sequence of the steel ring. The generation of seam feature maps includes: The image sequence of the steel ring surface is analyzed by performing image overlap analysis on adjacent image frames to determine multiple overlapping regions; The multiple overlapping regions are registered according to the image sequence of the steel ring surface to construct a surface unfolded map of the steel ring. Edge detection is performed based on the surface development diagram of the steel ring to extract the seam edge contour data; Based on the surface development diagram of the steel ring, a polar coordinate system is constructed, and the seam edge contour data is converted to the polar coordinate system to determine multiple seam contour data, which include radial distance data and angular position data. The joint feature map is generated by splicing the radial distance data and the angular position data together. The generation of a joint quality report includes: Construct a joint quality evaluation matrix, and perform weighted analysis on the multiple joint parameters based on the joint quality evaluation matrix to obtain multiple joint weight coefficients; The multiple joint parameters are integrated and analyzed according to the multiple joint weight coefficients to determine the comprehensive quality score; Based on the comprehensive quality score, the joint area is traversed for judgment, and multiple substandard joint areas are extracted. The multiple substandard joint areas contain multiple joint defect coordinates. Spatial clustering analysis was performed on the multiple substandard joint areas to obtain isolated defect classes and continuous defect classes. The coordinates of the multiple joint defects are mapped to the polar coordinate system according to the isolated defect class and the continuous defect class, thereby generating multiple defect location markers; Add the multiple defect location markers to the joint quality report.

2. The automatic seam detection method for electric tricycle steel rims based on image processing as described in claim 1, characterized in that, Morphological analysis based on the seam feature map is performed to identify multiple seam parameters. The method includes: Multi-scale edge enhancement processing is performed based on the seam edge contour data to extract edge contour pixel data of multiple seam regions. The seam feature map is traversed, and the radial distance data and angular position data are analyzed according to the edge contour pixel data to construct a data mapping relationship network. The seam morphology of the seam feature map is analyzed according to the data mapping relationship network to identify the multiple seam parameters.

3. The automatic seam detection method for electric tricycle steel rims based on image processing as described in claim 2, characterized in that, Multi-scale edge enhancement processing is performed based on the seam edge contour data to extract edge contour pixel data of multiple seam regions. The method includes: Based on the surface development diagram of the steel ring, a non-subsampled wavelet transform is performed to extract edge response features; Edge tracking is performed based on the edge response features to generate polygonal data of the seam contour. The gradient magnitude of the edge points is calculated based on the polygonal data of the seam contour. The seam edge contour data is then searched and filtered according to the gradient magnitude of the edge points to determine multiple valid edge points. The multiple effective edge points are connected at the pixel level to obtain the edge contour pixel data of multiple seam regions.

4. The automatic seam detection method for electric tricycle steel rims based on image processing as described in claim 3, characterized in that, The method involves performing seam morphology analysis on the seam feature map according to the data mapping network to identify the multiple seam parameters, including: Based on the polar coordinate system, the edge contour pixel data is transformed and detected to determine the seam centerline. Based on the seam centerline, deviation analysis is performed to generate an angle deviation value. A sliding window is set according to the data mapping relationship network, and the edge contour pixel data is synchronized to the sliding window for calculation to construct a straightness evaluation index. The average distance between the two edges of the joint is retrieved, and the average distance is calibrated to determine the joint width data; The straightness evaluation index and the joint width data are corrected according to the angle deviation value to obtain the multiple joint parameters.

5. The automatic seam detection method for electric tricycle steel rims based on image processing as described in claim 1, characterized in that, The method includes triggering a seam detection signal based on the seam quality report, and automatically detecting seams in the steel rim of the electric tricycle based on the seam detection signal. When there are N defect location markers in the steel rim of an electric tricycle, the N defect location markers are discretely analyzed to generate discretely distributed defect data. Based on the discrete distribution defect data, a composite defect early warning signal is triggered. The composite defect early warning signal includes a crack-finding signal. Based on the composite defect early warning signal, the defect is classified and a defect severity level is constructed. The defect severity level is used as an index to match the sorting strategy database for electric tricycle steel rims to generate a target sorting strategy. Based on the seam-finding signal, the target sorting strategy is executed to re-inspect and verify the defective seams, and the steel rims of the electric tricycle are automatically seam-finding detection and correction based on the verification results.

6. The automatic seam detection method for electric tricycle steel rims based on image processing as described in claim 5, characterized in that, The method involves using the defect severity level as an index to match the sorting strategy database for electric tricycle steel rims to generate a target sorting strategy, including: Based on the defect severity level, the sorting strategy database is matched to construct a rework priority level; According to the rework priority level, the coordinates of the multiple joint defects are synchronized to the automatic grinding workstation for recording, and a defect grinding file is generated. Based on the defect polishing files, a periodic analysis is performed to obtain the defect distribution pattern. The defect distribution pattern is then traced back to the sorting strategy database to select and determine the target sorting strategy.

7. The automatic seam detection method for electric tricycle steel rims based on image processing as described in claim 5, characterized in that, Based on the seam-finding signal, the target sorting strategy is executed to re-inspect and verify the defective seams. Based on the verification results, the steel rim of the electric tricycle is automatically detected and corrected. The method includes: Based on the seam-finding signal, reference markers are set in the non-seamless area of ​​the electric tricycle's steel rim, and the repeatability of the rotating platform is calculated based on the reference markers. The target sorting strategy is executed in multiple cycles to obtain multi-cycle sorting data. The multi-cycle sorting data is analyzed according to the repeatability of the rotating platform. Based on the analysis results, defective seams are evaluated, and system detection results are generated. The manual detection results are compared with the system detection results for verification, and the verification results are generated.

Images