Point cloud construction-based hook center line extraction method
By introducing multi-angle imaging and controllable lighting systems in hook point cloud construction, combined with efficient image processing and point cloud construction algorithms, the problem of low accuracy in hook point cloud construction and centerline extraction is solved, and high-precision and reliable hook three-dimensional point cloud construction and centerline extraction are achieved.
Patent Information
- Application Number
- CN202510000407.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-02
- Publication Date
- 2025-04-25
AI Technical Summary
The prior art has problems of low accuracy and insufficient reliability in hook point cloud construction and centerline extraction, especially due to the single color of the hook and the lack of obvious features on the surface, which leads to difficulty in extracting feature points and large measurement errors.
Multi-angle imaging and controllable lighting system are used to enhance the details and contrast in the image, feature points are extracted through SIFT algorithm, combined with RANSAC algorithm for matching and removal, high-precision three-dimensional point clouds are generated using triangulation and ICP algorithm, and the center line of the hook is extracted through circular fitting and B-spline fitting technology.
The accuracy and stability of feature point extraction are significantly improved, the generated hooked three-dimensional point cloud has high accuracy, and the extracted centerline has higher accuracy and smoothness, verifying the high accuracy and reliability of the method.
Smart Images
Figure CN119888262A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of computer vision and point cloud construction, and in particular to a method for extracting a hook centerline based on point cloud construction. Background Art
[0002] In industrial manufacturing and quality inspection, hooks are a common connection and support component. Their geometry and precision have a significant impact on the quality and performance of products. Accurately measuring the three-dimensional shape and centerline of hooks is of great significance to ensure product quality and performance.
[0003] Existing methods for constructing hook point clouds mainly rely on laser scanning and depth camera technology. Although laser scanning technology can provide high-precision three-dimensional data, its equipment cost is high, the operation is complex, and the environmental requirements are high, making it unsuitable for large-scale applications. Depth camera technology can obtain three-dimensional point cloud data more conveniently, but when processing objects with complex surface features and reflectivity, it is easy to generate noise and data loss, affecting the measurement accuracy.
[0004] The hook has a single color and lacks obvious features on the surface, which makes it difficult to extract stable feature points during image processing, resulting in low accuracy in point cloud construction. In addition, images taken at different angles may have errors in feature point matching and point cloud registration, further affecting the accuracy of the measurement results. The existing technology is still insufficient in solving these problems, so an improved method is urgently needed to improve the accuracy and reliability of hook point cloud construction and centerline extraction. Summary of the invention
[0005] In view of the shortcomings of the prior art, the present invention proposes a method for extracting the center line of a hook based on point cloud construction, aiming to solve the problem of difficulty in extracting feature points and measurement errors caused by the single color of the hook. The method enhances the details and contrast in the image by introducing multi-angle imaging and a controllable lighting system, thereby improving the accuracy of feature point extraction. Then, a high-precision three-dimensional point cloud of the hook is generated using an improved image processing and point cloud construction algorithm, and the center line of the hook is extracted by circular fitting and B-spline curve fitting techniques. Finally, the accuracy and reliability of the method are verified by comparing with the measurement results of the depth camera.
[0006] In order to solve the above technical problems, the technical solution of the present invention is:
[0007] A hook centerline extraction method based on point cloud construction includes the following steps:
[0008] Step 1, obtain the hook image and preprocess it;
[0009] Step 2: Use SIFT algorithm to extract feature points from the preprocessed image;
[0010] Step 3: Preliminary matching of the same feature points under different perspectives is performed through similarity measurement of feature point descriptors, and RANSAC algorithm is used to remove false matches;
[0011] Step 4: Calculate the three-dimensional coordinates of the feature points using the triangulation method to generate a preliminary three-dimensional point cloud, and register and merge the draft point clouds obtained from different angles using the ICP algorithm to generate a high-precision hook three-dimensional point cloud;
[0012] Step 5: Perform layered processing on the high-precision three-dimensional point cloud of the hook, perform circular fitting on the point set on each cross section, determine the center of the cross section, and extract the center line of the hook through the B-spline curve fitting method.
[0013] Preferably, in step 1, a multi-angle camera system is used to capture images of the hook from at least four different perspectives.
[0014] Preferably, the multi-angle camera system includes four high-resolution cameras, which are respectively arranged on the front, back, left and right sides of the hook.
[0015] Preferably, in step 1, when taking the hook image, a controllable lighting system is arranged around the hook to enhance the details and contrast in the image by adjusting the angle and brightness of the light source.
[0016] Preferably, the controllable lighting system includes four LED light sources with adjustable brightness and angle, which are arranged around the hook to form ring-shaped lighting.
[0017] Preferably, in step 1, the image preprocessing method includes using a Gaussian filter to remove noise and perform grayscale adjustment.
[0018] Preferably, in step 3, Euclidean distance is used as a similarity metric for feature point descriptors.
[0019] Preferably, in step 5, the circle fitting method is: using the least squares method to calculate the optimal center position and radius of each layer of point set.
[0020] Preferably, in step 5, the B-spline curve fitting method is:
[0021] Firstly, a set of control points representing the target hook shape is extracted from the 3D point cloud by using the curvature change rate, point density distribution and fitting error with the target shape as screening conditions;
[0022] Then, a piecewise cubic polynomial basis function is used to ensure the smoothness and continuity of the curve. The node vectors of the basis function are generated by uniform distribution or adaptive distribution according to the point cloud density to ensure that the fitting result is consistent with the spatial distribution of the control points.
[0023] Finally, a smooth centerline curve is generated by solving the B-spline equation using the extracted control points and the constructed basis functions, and the curve can accurately describe the center structure of the hook.
[0024] As a preferred method, accuracy verification is also included, and the specific method is as follows:
[0025] Use a depth camera to shoot the hook from different angles to generate three-dimensional point cloud data of the hook;
[0026] The point cloud data acquired by the depth camera is layered and circularly fitted to extract the center line of the hook;
[0027] The center lines obtained by the multi-angle imaging method and the depth camera method are compared, and the error between the two is calculated using point-to-point distance and curve fitting error indicators.
[0028] The present invention has the following characteristics and beneficial effects:
[0029] By adopting the above technical scheme, by introducing multi-angle imaging and controllable lighting system, the present invention can significantly improve the details and contrast in the hook image, thereby improving the accuracy of feature point extraction. Using the improved point cloud construction algorithm, a high-precision three-dimensional point cloud of the hook can be generated. The circular fitting and B-spline curve fitting technology ensure the accuracy of the hook centerline extraction. The high accuracy and reliability of the method are verified by comparison with the depth camera measurement results. The overall technical effects include: improving the accuracy and stability of feature point extraction; improving the accuracy of point cloud construction; the extracted hook centerline has higher accuracy and smoothness; verifying the feasibility and high accuracy of the method. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0031] Figure 1 This is a flowchart of a hook centerline extraction method based on point cloud construction according to an embodiment of the present invention.
[0032] Figure 2 for Figure 1 Flow chart of the centerline extraction method.
[0033] Figure 3 for Figure 1 Flow chart of the medium-precision verification method.
[0034] Figure 4 It is a schematic diagram of hierarchical processing in center line extraction of a hook center line extraction method based on point cloud construction in an embodiment of the present invention.
[0035] Figure 5 This is a schematic diagram of circular fitting in centerline extraction of a hook centerline extraction method based on point cloud construction in an embodiment of the present invention. DETAILED DESCRIPTION
[0036] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.
[0037] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.
[0038] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood by specific circumstances.
[0039] The present invention provides a hook centerline extraction method based on point cloud construction, such as Figure 1 As shown, the specific steps are as follows:
[0040] S1: Lighting Condition Adjustment
[0041] The controllable lighting system is introduced to enhance the details and contrast in the image by adjusting the angle and brightness of the light source. The system uses four LED light sources, which are arranged around the hook to form a ring lighting. By adjusting the brightness and angle of the light source, the best lighting effect is produced, thereby enhancing the image details and facilitating the extraction of feature points.
[0042] S2: Multi-angle shooting
[0043] A multi-angle camera system is used to capture the hook from at least four different angles to ensure comprehensive image information. The cameras are carefully positioned on the front, back, left and right sides of the hook to cover all key parts of the hook and avoid blind spots. Each camera is equipped with a high-resolution lens and autofocus function to ensure clear and detailed images.
[0044] S3: Image preprocessing
[0045] The images taken at each angle are preprocessed. First, a Gaussian filter is used to remove noise from the image. The Gaussian filter reduces random noise in the image through smoothing while maintaining the edge features of the image. Then, the image is grayscale adjusted to enhance the contrast and brightness of the image, making the feature points more prominent and easier to extract. Finally, the SIFT (Scale Invariant Feature Transform) algorithm is used to extract feature points from the image. The SIFT algorithm detects extreme points in the image and calculates the scale, direction, and position of these points to generate a set of feature descriptors with rotation, scale, and brightness invariance.
[0046] In this embodiment, the specific method of the SIFT algorithm is as follows:
[0047] Construction of scale space: By performing Gaussian filtering on the image, images of different scales are constructed to detect multi-scale features. The specific formula is:
[0048] L(x,y,σ)=G(x,y,σ)*I(x,y)
[0049] Where L(x, y, σ) represents the scale space image, G(x, y, σ) is the Gaussian kernel function, and I(x, y) is the original image.
[0050] Difference of Gaussian (DoG) calculation: In order to detect key points in an image, the difference between adjacent scale images is calculated:
[0051] D(x,y,σ)=L(x,y,kσ)-L(x,y,σ)
[0052] The feature areas in the image are emphasized by using Difference of Gaussian (DoG).
[0053] Key point detection and precise positioning: In the differential Gaussian image, detect local extreme points. Perform sub-pixel precise positioning of extreme points using the optimization formula:
[0054]
[0055] Among them, H is the Hessian matrix and g is the gradient vector.
[0056] Feature point description: Each key point has direction information, describing the gradient direction and magnitude around the point, which is used to generate feature descriptors. The gradient direction is calculated by the formula:
[0057]
[0058] S4: Feature point matching
[0059] Perform feature point matching on the preprocessed image. Preliminary matching finds possible matching pairs through similarity measurement of feature point descriptors. The similarity measurement uses Euclidean distance to calculate the difference between feature descriptors, and feature point pairs with smaller distance are considered as potential matching pairs.
[0060] Specifically, in this embodiment, the method of calculating the difference between feature descriptors using the Euclidean distance is as follows:
[0061] First, the transformation model is calculated (perspective transformation) and the error is calculated:
[0062]
[0063] Where e represents the Euclidean error between the transformed point and the true matching point. x1, y1 represent the pixel coordinates of the original point. Represents the point coordinates calculated by the perspective transformation matrix.
[0064] Then, internal point determination: if e < ∈, the matching pair is determined to be an internal point, otherwise it is an external point, where ∈ represents the error threshold, which is used to determine the standard for internal and external points.
[0065] Then, the RANSAC (Random Sampling Consistency) algorithm is used to remove false matches. The RANSAC algorithm randomly selects feature point pairs through iterations, calculates the transformation matrix, verifies all feature point pairs, and removes feature point pairs that are inconsistent with the transformation matrix. Finally, the three-dimensional coordinates of each feature point are calculated by triangulation to form a preliminary three-dimensional point cloud.
[0066] Specifically, the number of iterations of the RANSAC algorithm is expressed as follows:
[0067]
[0068] S5: Point cloud construction
[0069] Based on the feature point matching results, the ICP (Iterative Closest Point) algorithm is used to align and merge the point clouds acquired at different angles. The ICP algorithm minimizes the distance error between point clouds through an iterative process to ensure accurate alignment and fusion of point clouds. The point cloud data is fused and optimized, including removing isolated points, smoothing, and noise filtering. Removing isolated points deletes isolated points by setting the neighborhood radius; smoothing smoothes the point cloud through the moving least squares method; and noise filtering removes noise points in the point cloud through statistical filtering and conditional filtering. Finally, a high-precision hook 3D point cloud is generated.
[0070] In this embodiment, the specific implementation method is as follows:
[0071] First, point cloud matching
[0072] For each point t in the target point cloud T i , find the point s closest to it in the source point cloud S j , the matching pair is (t i ,s j ). Minimize the Euclidean distance between pairs of points:
[0073] d(t i ,s j )=|t i -s j |
[0074] where t i ∈T, is a point in the target point cloud, s j ∈S is a point in the source point cloud, d(t i ,s j ) is a point pair (t i ,s j ) between the two.
[0075] Then, calculate the transformation matrix
[0076] Calculate the optimal rigid transformation matrix (rotation matrix R and translation vector t) from the source point cloud S to the target point cloud T. The goal is to minimize the distance error between the transformed source point cloud and the target point cloud:
[0077]
[0078] Where R represents the rotation matrix, t represents the translation vector, and s i represents the i-th point in the source point cloud, t i Represents the i-th point in the target point cloud.
[0079] Finally, the least squares problem is solved
[0080] The minimization error function E(R, t) is solved by the least squares method, using the mean and covariance matrices of the aligned point pairs to calculate the rotation matrix R and the translation vector t.
[0081] The rotation matrix is calculated by singular value decomposition (SVD) or Kabsch algorithm, and the calculation formula is as follows:
[0082]
[0083] Where R represents the rotation matrix, t represents the translation vector, and s i represents the i-th point in the source point cloud, t i Represents the i-th point in the target point cloud.
[0084] The translation vector is obtained by the mean calculation, and the calculation formula is as follows:
[0085]
[0086] Where N is the number of points, R represents the rotation matrix, t represents the translation vector, s i represents the i-th point in the source point cloud, t i Represents the i-th point in the target point cloud.
[0087] S6: Centerline Extraction
[0088] The 3D point cloud data is processed in layers, with each layer corresponding to a cross section of the hook. Figure 2 As shown, the specific steps are as follows:
[0089] S601: Layered processing
[0090] Specifically, Figure 4 As shown, the rotated workpiece image is gray-scaled to obtain a single-channel workpiece image with only gray-scale values from the original RGB three-channel. Then, the cross-section is divided along the main axis of the hook using a fixed-interval slicing method, and each slice corresponds to a plane perpendicular to the main axis.
[0091] S602: Circular Fitting
[0092] Specifically, Figure 5 As shown, a circular fitting is performed on each cross-sectional point set to determine the cross-sectional center of the hook. The fitting process uses the least squares method to calculate the optimal center position and radius of each layer of point set.
[0093] S603: Curve Fitting
[0094] All the center lines are smoothed by B-spline curve fitting technology. B-spline curve fitting generates a smooth center line curve through control points and basis functions to ensure the continuity and smoothness of the center line. The method of B-spline curve fitting is:
[0095] Firstly, by analyzing the geometric features of the point cloud data, a set of control points representing the target hook shape is extracted from the point cloud using the curvature change rate, point density distribution and fitting error with the target shape as screening conditions;
[0096] Then, a piecewise cubic polynomial basis function is used to ensure the smoothness and continuity of the curve. The node vectors of the basis function are generated by uniform distribution or adaptive distribution according to the point cloud density to ensure that the fitting result is consistent with the spatial distribution of the control points.
[0097] Comparative Example
[0098] S7: Accuracy Verification
[0099] The same point cloud is constructed for the hook using the depth camera. Figure 3 As shown, the specific steps are as follows:
[0100] S701: Depth Camera Measurement
[0101] The hook is photographed from different angles by a depth camera to generate 3D point cloud data of the hook. The depth camera has autofocus and high-resolution capture functions to ensure the acquisition of high-precision 3D point cloud.
[0102] S702: Centerline Extraction
[0103] The point cloud data acquired by the depth camera is layered and circularly fitted to extract the center line of the hook.
[0104] S703: Error Analysis
[0105] The center lines obtained by the multi-angle imaging method and the depth camera method are compared. The error between the two is calculated to evaluate the measurement accuracy of the method of the present invention. The error calculation uses indicators such as point-to-point distance and curve fitting error to ensure that the measurement accuracy is fully and accurately reflected. The experiment shows that the error between the center line extracted in this example and the high-precision depth camera is less than the threshold, achieving a result close to that of the depth camera at a lower economic cost.
[0106] Through the above steps, the present invention realizes high-precision and high-efficiency hook point cloud construction and centerline extraction, and verifies the accuracy of the method by comparison with a depth camera. The results show that the method of the present invention has high measurement accuracy and reliability, and provides a reliable technical means for hook detection in industrial manufacturing and quality control.
[0107] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions and variations of these embodiments including components are made without departing from the principles and spirit of the present invention, and still fall within the scope of protection of the present invention.
Claims
1. A hook centerline extraction method based on point cloud construction, characterized in that: The steps include: Step 1, obtain the hook image and pre-process it; Step 2: Use SIFT algorithm to extract feature points from the preprocessed image; Step 3: Preliminary matching of the same feature points under different perspectives is performed through similarity measurement of feature point descriptors, and RANSAC algorithm is used to remove false matches; Step 4: Calculate the three-dimensional coordinates of the feature points using the triangulation method to generate a preliminary three-dimensional point cloud, and register and merge the draft point clouds obtained from different angles using the ICP algorithm to generate a high-precision hook three-dimensional point cloud; Step 5: Perform layered processing on the high-precision three-dimensional point cloud of the hook, perform circular fitting on the point set on each cross section, determine the center of the cross section, and extract the center line of the hook through the B-spline curve fitting method.
2. The hook centerline extraction method based on point cloud construction according to claim 1 is characterized in that: In step 1, a multi-angle camera system is used to capture hook images from at least four different locations.
3. The hook centerline extraction method based on point cloud construction according to claim 2 is characterized in that: The multi-angle camera system includes four high-resolution cameras, which are respectively arranged on the front, back, left and right sides of the hook.
4. The hook centerline extraction method based on point cloud construction according to claim 1 is characterized in that: In step 1, when taking the hook image, a controllable lighting system is arranged around the hook to enhance the details and contrast in the image by adjusting the angle and brightness of the light source.
5. The hook centerline extraction method based on point cloud construction according to claim 4 is characterized in that: The controllable lighting system includes four LED light sources with adjustable brightness and angle, which are arranged around the hook to form ring lighting.
6. The method for extracting hook centerline based on point cloud construction according to claim 1, characterized in that: In step 1, the image preprocessing method includes using a Gaussian filter to remove noise and perform grayscale adjustment.
7. The method for extracting hook centerline based on point cloud construction according to claim 1, characterized in that: In step 3, the Euclidean distance is used as a similarity metric for feature point descriptors.
8. The method for extracting hook centerline based on point cloud construction according to claim 1, characterized in that: In step 5, the circle fitting method is: using the least square method to calculate the optimal center position and radius of each layer of point set.
9. The method for extracting hook centerline based on point cloud construction according to claim 1, characterized in that: In step 5, the B-spline curve fitting method is: Firstly, a set of control points representing the target hook shape is extracted from the 3D point cloud by using the curvature change rate, point density distribution and fitting error with the target shape as screening conditions; Then, a piecewise cubic polynomial basis function is used to ensure the smoothness and continuity of the curve. The node vectors of the basis function are generated by uniform distribution or adaptive distribution according to the point cloud density to ensure that the fitting result is consistent with the spatial distribution of the control points. Finally, a smooth centerline curve is generated by solving the B-spline equation using the extracted control points and the constructed basis functions, and the curve can accurately describe the center structure of the hook.
10. The method for extracting hook centerline based on point cloud construction according to claim 1, characterized in that: It also includes accuracy verification, the specific method is: Use a depth camera to shoot the hook from different angles to generate three-dimensional point cloud data of the hook; The point cloud data acquired by the depth camera is layered and circularly fitted to extract the center line of the hook; The center lines obtained by the multi-angle imaging method and the depth camera method are compared, and the error between the two is calculated using point-to-point distance and curve fitting error indicators.