Three-dimensional point cloud-based bead toe edge type defect detection method

Abstract

The invention discloses a bead toe edge type defect detection method based on three-dimensional point cloud, and relates to the technical field of computer vision, three-dimensional point cloud processing and automatic quality detection.The bead toe edge type defect detection method comprises the steps that three-dimensional point cloud data of the surface of a tire is obtained through a non-contact three-dimensional scanning technology; preprocessing the three-dimensional point cloud data through a filtering algorithm and down-sampling; segmenting the preprocessed three-dimensional point cloud data based on a region growing algorithm, and setting different distances and normal vector thresholds for irrelevant regions by setting local stop conditions to obtain a segmentation result of the three-dimensional point cloud data; and according to a segmentation result of the three-dimensional point cloud data, detecting and identifying bead toe edge defects by setting a corresponding matching rule, and generating a defect report. Therefore, by adopting the bead toe edge type defect detection method based on the three-dimensional point cloud, the bead toe edge type defects can be accurately and stably detected under the conditions of complex surfaces and different tire types, and the detection efficiency and accuracy are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical fields of computer vision, three-dimensional point cloud processing and automated quality inspection, and in particular to a method for detecting tire toe-out edge defects based on three-dimensional point clouds. Background Art

[0002] In modern tire manufacturing and use, the bead toe area is a core component for ensuring tire performance and safety. As the critical area where the tire contacts and seals with the rim, the quality of the bead toe directly impacts tire assembly stability, sealing performance, and overall vehicle safety. However, during production and use, the bead toe area is often susceptible to defects such as oversize, material overflow, or abnormal shape due to complex process pressures, improper material flow control, and various environmental factors. These issues not only affect the tire's appearance, leading to reduced strength, sealing, and driving performance, but also pose a potential threat to vehicle safety. Therefore, detecting bead toe defects has become a crucial step in ensuring tire quality and safety.

[0003] Currently, common methods for detecting tire toe defects include manual inspection and two-dimensional image recognition technology. Both methods have played a certain role in practice, but they also have many limitations. Manual inspection is the most traditional detection method, relying primarily on the operator's experience. It inspects the tire toe area through visual observation or using simple tools to determine whether there are problems such as material overflow, excessive size, or abnormal shape. Due to its simplicity and low cost, it was widely used in early tire manufacturing. However, manual inspection requires high operator expertise and judgment, and is easily affected by human subjective factors such as fatigue, lack of experience, and interference from ambient light, resulting in inconsistent detection results. In addition, manual inspection is inefficient and cannot meet the needs of large-scale inspection in assembly line production. Especially in high-intensity work environments, it is prone to misjudgments or omissions, and cannot guarantee quality monitoring throughout the entire process.

[0004] Two-dimensional image recognition technology uses a camera to capture a flat image of the tire toe area and utilizes image processing algorithms to identify and analyze defects. This method can improve detection efficiency and automation to a certain extent, gradually replacing some manual inspections in modern tire production. Two-dimensional image recognition can quickly analyze the shape and size of the tire toe, and is particularly well-suited for simple planar inspection tasks. However, this technology also has significant limitations. Because two-dimensional images only provide planar perspective information, recognition accuracy is low for tire toes with complex curvatures, irregular shapes, or hidden areas, making it difficult to fully capture detailed defects. Furthermore, lighting conditions and surface reflectivity can significantly affect image quality, limiting the stability and reliability of detection.

[0005] Overall, while manual inspection and 2D image recognition have played a role in tire toe defect detection, with the increasing complexity of tire design and the increasing precision required in manufacturing, these two traditional methods can no longer fully meet the demands of modern production for high-precision, high-efficiency, and comprehensive inspection. More advanced 3D inspection technologies and intelligent algorithms are urgently needed to address the shortcomings of current technologies and provide more reliable quality assurance for the tire manufacturing industry. Summary of the Invention

[0006] The purpose of the present invention is to provide a method for detecting tire toe edge defect based on three-dimensional point cloud, which can accurately and quickly identify and locate tire toe edge defect.

[0007] To achieve the above object, the present invention provides a method for detecting tire toe-out edge defects based on a three-dimensional point cloud, comprising the following steps: S1. Use non-contact 3D scanning technology to obtain 3D point cloud data of the tire surface, and pre-process the 3D point cloud data through filtering algorithms and downsampling; S2. Segment the preprocessed 3D point cloud data based on the region growing algorithm, and obtain the segmentation result of the 3D point cloud data by setting local stopping conditions and different distance and normal vector thresholds for irrelevant regions; S3. Based on the segmentation results of the 3D point cloud data, by setting corresponding matching rules, detect and identify tire toe edge defects and generate a defect report.

[0008] Preferably, in step S1, three-dimensional point cloud data of the tire is collected in all directions through multi-angle scanning or rotation scanning.

[0009] Preferably, in step S1, the filtering algorithm includes median filtering and mean filtering.

[0010] Preferably, in step S1, a voxel grid-based downsampling method is used to uniformly sample the three-dimensional point cloud data.

[0011] Preferably, in step S2, the region growing algorithm includes the following steps: S21. Based on curvature analysis and normal vector analysis, identify characteristic areas on the tire surface, and determine seed points based on the geometry of the tire toe area; S22. Based on the similarity of the distance and normal vector between two points, expand from the seed point to the neighboring points, and set local stopping conditions through different distance thresholds and normal vector thresholds to segment different geometric areas.

[0012] Preferably, in step S21, the fetal toe portion is preliminarily determined by performing curvature analysis on the preprocessed point cloud, and then the geometric features of the fetal toe portion are obtained by normal vector analysis, and then the seed point is selected.

[0013] Preferably, in step S22, the distance threshold is set according to the point cloud density and the geometric characteristics of the tire toe area, and the normal vector threshold is set based on the statistical distribution of the curvature change and the normal vector angle.

[0014] Preferably, in step S3, the basis for detecting the tire toe edge defect includes abnormal curvature change, edge irregularity and regional morphological variation.

[0015] Therefore, the present invention adopts the above-mentioned tire toe edge defect detection method based on three-dimensional point cloud, which has the following technical effects: (1) The present invention uses three-dimensional point cloud technology for defect detection. Compared with the traditional two-dimensional image method, it can provide richer spatial information, overcome the problems of tire surface curvature change and texture complexity, and can adapt to tires of different models and specifications. It can perform customized detection based on the different forms and characteristics of the tire surface, accurately identify the complex geometric forms of the tire toe area (such as tread pattern and sidewall curvature change), thereby achieving high-precision identification of tire toe edge type defects, greatly improving the accuracy of detection, strong adaptability, and has broad application prospects.

[0016] (2) The region growing algorithm used in this invention determines the initial seed point based on characteristic regions of the tire surface (such as the geometry of the tire toe area). Compared with the random selection of seed points in conventional region growing algorithms, the selection of seed points based on geometric features ensures that growth begins at the key location of the tire toe area, more accurately capturing the geometric features of the tire toe area. This avoids segmentation errors that may result from random selection of seed points, ensures segmentation accuracy and specificity, and thus improves the accuracy of defect detection.

[0017] (3) The present invention sets a local stop condition based on the geometric features of the tire toe area, and sets different distance and normal vector thresholds according to the different geometric forms of the tire toe area. This can more flexibly adapt to the complex geometric forms of the tire toe area, avoid the problem of over-segmentation or under-segmentation, and improve the accuracy and stability of the segmentation results. At the same time, combined with the similarity of normal vectors, it can more accurately judge the geometric relationship between points. Compared with a single distance threshold setting, it can more accurately segment areas with similar geometric forms, especially when processing complex curved surfaces, showing better robustness and improving the detection accuracy of tire toe edge defects.

[0018] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1The present invention is a tire detection result in an embodiment of a tire toe-out defect detection method based on a three-dimensional point cloud, wherein (a) is a tire defect-free detection result, and (b) is a tire toe-out defect detection result; Figure 2 The unprocessed three-dimensional point cloud data in an embodiment of a tire toe edge defect detection method based on a three-dimensional point cloud; Figure 3 The present invention is a 3D point cloud data pre-processed in an embodiment of a tire toe edge defect detection method based on a 3D point cloud; Figure 4 A flow chart of a region growing algorithm in an embodiment of a tire toe edge defect detection method based on a three-dimensional point cloud; Figure 5 The present invention is a flow chart of a defect determination algorithm in an embodiment of a tire toe edge defect detection method based on three-dimensional point cloud. DETAILED DESCRIPTION

[0020] The present invention can be explained in more detail by the following examples. The purpose of disclosing the present invention is to protect all changes and improvements within the scope of the present invention. The present invention is not limited to the following examples.

[0021] A bead toe defect is a common surface defect during tire production. It typically manifests as an irregular edge at the junction of the tread and sidewall, potentially reducing the tire's strength, sealing, and driving performance. This defect not only affects the tire's appearance but can also have a serious impact on tire safety. However, existing defect detection technologies have the following technical issues when detecting bead toe defects: Limitations of Traditional Inspection Methods: Existing tire toe defect detection relies primarily on manual inspection or 2D image recognition technology. Manual inspection is time-consuming and susceptible to human error. Furthermore, 2D images lack sufficient spatial information to accurately identify subtle defects at the tire toe. Existing methods often struggle to achieve comprehensive and accurate detection, especially when the tire surface has curvature or irregular shapes.

[0022] Insufficient detection capability for complex bead defects: Bead defects are complex in shape and may appear as tiny bumps, depressions, or irregular edges. Traditional detection methods are often unable to accurately identify these subtle and localized defects, and are prone to missed or false detections.

[0023] Poor adaptability to tire surface variability: Due to the complex tire production process, the surface morphology and structure of the tire toe area often change. Traditional detection methods have poor adaptability to changes in tire surface materials, tread patterns, and surface deformation, which can easily lead to unstable detection and affect the accuracy and consistency of the test results.

[0024] Lack of efficient automated detection methods: Currently, the detection of tire toe defects mostly relies on manual intervention or simple image recognition algorithms. There is a lack of efficient and accurate automated detection methods, which cannot meet the requirements of large-scale production lines for detection speed, accuracy and consistency.

[0025] To solve the above problems, the present invention provides a method for detecting tire toe edge defects based on three-dimensional point clouds, comprising the following steps: S1. Utilizing non-contact 3D scanning technology, laser scanners, structured light sensors, or other advanced 3D imaging devices are used to acquire high-precision 3D point cloud data of the tire surface. This scanning system boasts high spatial resolution, accurately capturing subtle variations in the tire's surface, including details in the tire toe area. Multi-angle scanning or rotational scanning ensures comprehensive 3D data capture of the tire, preventing defects that might be missed due to viewing angle limitations.

[0026] To ensure the quality of point cloud data, the present invention also introduces a variety of data acquisition optimization strategies, such as selecting an appropriate scanning frequency and adjusting the scanning path and angle to reduce the impact of factors such as lighting and differences in tire surface material.

[0027] Since the scan data may contain noise, redundant points and incomplete data, especially in the area with complex curvature changes on the tire surface (such as Figure 2 As shown in the figure, these noise points may interfere with the segmentation effect of the region growing algorithm and lead to incorrect segmentation results. Therefore, it is very important to preprocess the collected 3D point cloud data (such as noise removal and point cloud registration) to ensure the robustness of the subsequent region growing algorithm.

[0028] The preprocessing specifically includes: Noise removal: Use filtering algorithms such as median filtering and mean filtering to remove noise points during the scanning process and retain valid data.

[0029] Point cloud downsampling: In order to improve processing efficiency and reduce computational burden, this paper uses a voxel grid-based downsampling method to uniformly sample the point cloud, reduce the point cloud density, and maintain the overall morphological characteristics of the point cloud, such as Figure 3 shown.

[0030] The above preprocessing effectively removes noise points from the point cloud, reduces point cloud density, and reduces computational burden, while maintaining the overall morphological characteristics of the point cloud. This ensures data quality and usability, providing a high-quality data foundation for subsequent seed point selection. Furthermore, the method effectively addresses the limitations of traditional methods in complex environments, such as the impact of factors like tire surface stains and lighting variations on detection accuracy. This makes the method applicable to diverse environmental conditions, ensuring the reliability and consistency of detection results, and demonstrating strong robustness and anti-interference capabilities.

[0031] S2, such as Figure 4 As shown, the preprocessed point cloud data is segmented using the region growing algorithm, leveraging point-to-point similarity. Due to the complex curvature and geometry of the tire surface, especially the tire toe area, which may contain irregular protrusions and depressions, conventional region growing algorithms typically use fixed global stopping conditions (such as fixed distance thresholds or normal vector thresholds), which cannot meet the requirements for defects such as tire toe edge defects.

[0032] In order to adapt the region growing algorithm to this complex surface structure, it is necessary to ensure that the geometric features of the tire toe area can be accurately captured during the segmentation process, as follows: S21. The effectiveness of the region growing algorithm is highly dependent on the selection of initial seed points. Improper seed point selection may lead to incorrect region growing directions, thus affecting the accuracy of the segmentation results. Especially in the tire toe area, seed point selection needs to be based on characteristic regions of the tire surface (such as the geometry of the tire toe area) to ensure that the algorithm can accurately segment the target area. For this reason, this embodiment uses seed point selection based on geometric features, as follows: First, within the preprocessed point cloud data, characteristic regions of the tire surface are identified, particularly the bead toe area. Because the bead toe area typically exhibits specific geometric features (such as the junction with the tread and sidewall), it can be identified through curvature analysis and normal vector analysis. Curvature analysis calculates the curvature of each point in the point cloud to identify areas with significant curvature variation, preliminarily determining the bead toe area. Normal vector analysis calculates the normal vector of each point in the point cloud to identify areas with significant normal vector variation and determine the geometric characteristics of the bead toe area. These analyses enable accurate location of the bead toe area, providing a basis for seed point selection.

[0033] Then, based on the recognition results of the feature regions, seed points are selected. There are two ways to select seed points: random selection, which involves randomly selecting a point from the point cloud as a seed point; and feature region-based selection, which involves selecting representative points based on the geometry of the fetal toe area. In practice, feature region-based selection is more common because it ensures that the seed points are located in key areas of the fetal toe area, thereby improving the accuracy and specificity of the region growing algorithm. The selection of seed points is crucial because it determines the direction and extent of region growth, directly affecting the accuracy of the segmentation results.

[0034] Selecting seed points through the above steps not only improves the segmentation accuracy of the region growing algorithm but also effectively avoids segmentation errors caused by improper seed point selection. This feature-region-based seed point selection method, combined with preprocessing and feature region recognition technology, significantly improves the accuracy and efficiency of tire toe edge defect detection, laying a solid foundation for subsequent defect detection and identification.

[0035] S22. Starting from the initial seed point, based on the distance between the two points and the similarity of the normal vector, gradually expand to the neighboring points. If the distance between the neighboring point and the current point is less than the set threshold and the normal vector is similar, the neighboring point will be added to the current region.

[0036] In the region growing algorithm, normal vector similarity is one of the important conditions for determining whether a neighboring point should be added to the current region. Normal vector similarity is usually determined by calculating the normal vector angle between two points. Specifically, the normal vector angle of two given points is calculated using the dot product formula. If the angle is less than the set threshold, the normal vectors of the two points are considered to be similar. In practical applications, the threshold of normal vector similarity needs to be adjusted according to the specific application scenario. For tire toe edge type defect detection, due to the complex geometry of the tire toe area and the large range of normal vector variation, a reasonable threshold needs to be set to ensure that the algorithm can accurately capture the geometric features of the tire toe area. This threshold can be determined through experiments, and is set to 10° in this embodiment.

[0037] By combining normal vector similarity with a distance threshold, the region growing algorithm can more accurately segment regions with similar geometry, such as the tire toe. This method not only effectively avoids grouping unrelated regions into the same region, but also improves the accuracy and stability of the segmentation results, demonstrating greater robustness when processing complex surfaces, providing a reliable foundation for subsequent defect detection and identification.

[0038] Conventional region growing algorithms typically use a single distance or normal vector threshold. However, tire toe defects can manifest as geometric variations of varying scales (such as tiny bumps or depressions). A single threshold setting makes it difficult to detect these small bumps, depressions, or irregular edges. Therefore, this embodiment sets reasonable local stopping conditions for the region growing algorithm to prevent overgrowth or the incorrect grouping of unrelated regions into the same region. This multi-threshold approach better adapts to the complex geometric variations of the tire toe region, ensuring that the algorithm can accurately segment regions of varying shapes.

[0039] When setting the local stopping conditions, different distance and normal vector thresholds are set according to the geometric characteristics of the tire toe area, so that the region growing algorithm can handle multi-scale segmentation tasks to ensure that both macro defect features can be captured and micro anomalies can be identified.

[0040] The distance threshold is set according to the point cloud density and the geometric characteristics of the tire toe area: When point cloud density is high, the distance threshold is set lower to capture subtle changes; when density is low, the threshold is appropriately increased to improve efficiency. Furthermore, complex geometry in the tire toe area (such as edges and flat areas) also requires different distance thresholds, with smaller thresholds in edge areas and larger thresholds in flat areas.

[0041] The normal vector threshold is set based on the statistical distribution of curvature changes and normal vector angles: In areas with large curvature changes (such as the tire toe edge), the normal vector threshold is set to a small value to capture subtle changes; in areas with smaller curvature changes, the threshold is appropriately increased. By statistically analyzing the distribution of normal vector angles in the point cloud, a reasonable threshold range can be set.

[0042] The aforementioned multi-threshold setting method enables the region growing algorithm to stop region expansion at the appropriate time (e.g., when the number or distance of neighboring points exceeds a set threshold). This effectively avoids overgrowth or the erroneous grouping of unrelated regions into the same region. This enables precise segmentation of complex geometric features in the tire toe region, improves the detection of tire toe defects, and demonstrates greater adaptability. Furthermore, these threshold setting methods can be optimized through experimental adjustments to further ensure the algorithm's ability to accurately segment the tire toe region, avoid overgrowth or erroneous segmentation, and enhance detection accuracy and efficiency.

[0043] S3, such as Figure 5 As shown in the figure, according to the segmentation results of the 3D point cloud data, the threshold setting is matched with the defect judgment rule to determine whether the current area is a tire toe defect and assign a color to the target defect, as shown in the figure. Figure 1 The red part shown in (b) separates the tire body from the toe defect.

[0044] In another embodiment, the present invention provides a 3D point cloud-based tire toe defect detection method that seamlessly integrates with automated production lines. This method uses automated equipment to perform real-time, high-speed tire scanning and defect detection, reducing the need for manual intervention and processing more data in a shorter time, significantly improving production line efficiency. Furthermore, automated testing eliminates errors caused by human factors, significantly reducing labor costs and ensuring the accuracy and consistency of test results.

[0045] Automated equipment also enables online inspection during tire production, identifying and correcting defects in real time during the production process, thereby reducing scrap and rework rates and lowering production costs. It also quickly and accurately detects, classifies, and reports tire toe defects, improving tire quality control, ensuring consistent product quality, and boosting the overall efficiency and competitiveness of the production line.

[0046] In another embodiment, the present invention uses a region growing algorithm and defect feature extraction technology to quickly and accurately locate tire toe defects and automatically generate a defect report. This report not only contains information such as the defect type, location, and severity, but also provides the defect's geometric characteristics, helping production line personnel take timely remedial measures and reducing the complexity of quality control.

[0047] Therefore, the present invention adopts the above-mentioned tire toe edge defect detection method based on three-dimensional point cloud, which can accurately and stably detect tire toe edge defect under complex surfaces and different tire shapes, significantly improving detection efficiency and accuracy.

[0048] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solutions of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for detecting tire toe-out defects based on three-dimensional point clouds, characterized in that: The following steps are involved: S1. Use non-contact 3D scanning technology to obtain 3D point cloud data of the tire surface, and pre-process the 3D point cloud data through filtering algorithms and downsampling; S2. Segment the preprocessed 3D point cloud data based on the region growing algorithm, and obtain the segmentation result of the 3D point cloud data by setting local stopping conditions and different distance and normal vector thresholds for irrelevant regions; S3. Based on the segmentation results of the 3D point cloud data, by setting corresponding matching rules, detect and identify tire toe edge defects and generate a defect report.

2. The method for detecting tire toe-out edge defects based on three-dimensional point cloud according to claim 1, characterized in that: In step S1, three-dimensional point cloud data of the tire is collected in all directions through multi-angle scanning or rotation scanning.

3. The method for detecting tire toe-out edge defects based on three-dimensional point cloud according to claim 1, characterized in that: In step S1, the filtering algorithm includes median filtering and mean filtering.

4. The method for detecting tire toe-out edge defects based on three-dimensional point cloud according to claim 1, characterized in that: In step S1, a voxel grid-based downsampling method is used to uniformly sample the three-dimensional point cloud data.

5. The method for detecting tire toe-out edge defects based on three-dimensional point cloud according to claim 1, characterized in that: In step S2, the region growing algorithm includes the following steps: S21. Based on curvature analysis and normal vector analysis, identify characteristic areas on the tire surface, and determine seed points based on the geometry of the tire toe area; S22. Based on the similarity of the distance and normal vector between two points, expand from the seed point to the neighboring points, and set local stopping conditions through different distance thresholds and normal vector thresholds to segment different geometric areas.

6. The method for detecting tire toe-out edge defects based on three-dimensional point cloud according to claim 5, characterized in that: In step S21, the fetal toe position is preliminarily determined by performing curvature analysis on the preprocessed point cloud, and then the geometric features of the fetal toe position are obtained through normal vector analysis, and then the seed point is selected.

7. The method for detecting tire toe-out edge defects based on three-dimensional point cloud according to claim 5, characterized in that: In step S22 , the distance threshold is set according to the point cloud density and the geometric characteristics of the tire toe area, and the normal vector threshold is set based on the statistical distribution of the curvature change and the normal vector angle.

8. The method for detecting tire toe-out edge defects based on three-dimensional point cloud according to claim 1, characterized in that: In step S3, the basis for detecting the tire toe edge defect includes abnormal curvature change, edge irregularity and regional morphological variation.

Images