Contour extraction method based on point cloud data
Through the contour extraction method based on point cloud data, RANSAC and Gaussian core technology is used to solve the problems of massive data and high energy consumption in high-precision point cloud data processing, more efficient data processing and real-time computing are achieved, and production processes are optimized. cost.
Patent Information
- Application Number
- CN202411054765.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-11-19
AI Technical Summary
The prior art faces the problems of massive data processing burden and high energy consumption in high-precision point cloud data processing, especially in large-scale industrial scenarios and digital twin modeling, resulting in increased costs and inefficiency in production processes.
The contour extraction method based on point cloud data is adopted, local planes are fitted through the RANSAC method, point cloud data is projected, noise points are eliminated using Gaussian kernel, point cloud edge probability is calculated, boundary points are judged, and contour points of the target point cloud are formed.
It effectively reduces the computing volume and energy consumption of data processing, optimizes the production process, reduces costs, and improves real-time computing capabilities in large-scale industrial scenarios.
Smart Images

Figure CN118982552A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of geographic information technology, and more specifically to a contour extraction method based on point cloud data. Background Art
[0002] With the rise of concepts such as "digital city", point cloud data has become an indispensable and important resource in today's society. Buildings are the most important components of cities. Based on remote sensing data, high-precision and rapid extraction of buildings and their contours can be achieved, which will provide important basic data for "digital cities". It is also one of the difficulties in the current research field.
[0003] Point cloud data refers to a set of vectors in a three-dimensional coordinate system.
[0004] At present, in the process of contour extraction of point cloud data, the laser radars on the market may face the following disadvantages when facing large-scale industrial scenes or digital twin modeling scenes:
[0005] ① Under the premise of high precision, LiDAR usually collects massive 3D point cloud data, which increases the burden of data processing;
[0006] ② When processing data, expensive and ultra-large-capacity floating-point computing devices are required, which will generate a lot of energy consumption.
[0007] Therefore, in view of this, the existing structure is studied and improved, and a contour extraction method based on point cloud data is provided, in order to achieve a more practical purpose. Summary of the invention
[0008] 1. Technical issues to be solved
[0009] In response to the problems existing in the prior art, the purpose of the present invention is to provide a contour extraction method based on point cloud data, which can assist in the real-time calculation of target objects in large industrial scenes, allowing business owners to obtain more and faster production data, thereby optimizing the production process, reducing costs, and promoting the transformation to intelligent manufacturing.
[0010] 2. Technical solution
[0011] To solve the above problems, the present invention adopts the following technical solutions.
[0012] A contour extraction method based on point cloud data, the specific implementation steps of the contour extraction method are as follows:
[0013] Step 1: Get the point cloud data, traverse each point of the target point cloud data through the traversal method, and then use the RANSAC method to fit the local plane of the k-neighborhood of the data point coordinates;
[0014] Step 2: The traversed point cloud data set is projected onto the fitted local plane as the original data points;
[0015] Step 3: Use the Gaussian kernel to eliminate the influence of noise points, perform noise reduction processing, and form a new local plane;
[0016] Step 4: Calculate the probability that each point is an edge point, and calculate the average of all probability groups;
[0017] Step 5: According to the average value of the probability group, the points with values less than the average value are boundary points, and those with values greater than the average value are internal points;
[0018] Step 6: According to the distribution of the determined boundary points and internal points, the result is obtained. The set of all boundary points constitutes the contour points of the target point cloud.
[0019] Furthermore, in the step 1, when acquiring the point cloud data, the point cloud data is acquired by performing data acquisition through a three-dimensional laser scanner.
[0020] Furthermore, in step 1, the specific implementation steps of the RANSAC method are as follows:
[0021] Step S1.1, randomly select a subset from the original data set as the hypothesized inliers;
[0022] Step S1.2, estimating model parameters based on the subset;
[0023] Step S1.3, traverse all data in the data set except the subset, if the data point is within the given error, mark it as an inlier, otherwise mark it as an outlier;
[0024] Step S1.4, all the inliers form a consistent set. If the number of points in the consistent set meets a given threshold, all the inliers in the consistent set are used to re-estimate the model parameters, and then the algorithm ends;
[0025] Step S1.5: If the number of inliers in the consistent set is less than the threshold, a new subset is selected and steps S1.1 to S1.4 are repeated;
[0026] Step S1.6: After iterations, select a consistent set with the largest number of inliers, use all the inliers in the consistent set to re-estimate the model parameters, and then terminate the algorithm.
[0027] Furthermore, in the step 1, a k-neighborhood search is performed in the local plane of the k-neighborhood to identify and remove noise points, and the k-neighborhood of the data point is analyzed to determine whether it is a noise point and perform corresponding processing.
[0028] Furthermore, in step 2, the projection of the point cloud data onto the local plane is completed, and the specific steps are as follows:
[0029] Step S2.1, loading point cloud data from a file;
[0030] Step S2.2, prepare camera parameters to describe the position and direction of the camera in three-dimensional space;
[0031] Step S2.3: perform projection calculation, normalize the calculated coordinates to obtain pixel coordinates.
[0032] Furthermore, in step S2.2, the camera's intrinsic parameter matrix K is a 3×3 matrix, and the camera's extrinsic parameter matrix is:
[0033] T=[R,t]
[0034] Where R is a special orthogonal matrix and t is a three-dimensional vector.
[0035] Furthermore, in step S2.3, for each point p in the point cloud, the calculation formula of its coordinates in the camera coordinate system is:
[0036]
[0037] in, are the homogeneous coordinates of the point p.
[0038] Furthermore, in step 3, when the Gaussian kernel eliminates noise points, the mathematical model formula for calculation is:
[0039]
[0040] Where x and x′ are data points, is the standard deviation, and ||xx′|| 2 is the square of the Euclidean distance.
[0041] 3. Beneficial effects
[0042] Compared with the prior art, the advantages of the present invention are:
[0043] ① This solution uses a 3D laser scanner to collect data and obtain point cloud data. Each point of the target point cloud data is traversed by the traversal method. The local plane of the k-neighborhood of the point is fitted using the RANSAC method. The point is then projected onto the local plane. The influence of noise points is eliminated using the Gaussian kernel. The probability of each point being an edge point is calculated, and the average of all probabilities is calculated. Points less than the average are boundary points, and points greater than the average are internal points. All boundary point sets constitute the contour points of the target point cloud. This calculation method effectively reduces the 3D data in industrial scenes, effectively reduces the amount of calculation during data processing, reduces the burden of data processing, and effectively preserves the main contours and features of the target 3D point cloud.
[0044] ② This solution improves the efficiency of data processing by reducing the amount of calculation during data processing. At the same time, when facing real-time calculations of target objects in large industrial scenarios, business owners can obtain more and faster production data, thereby optimizing production processes, reducing costs, and reducing data processing energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] Figure 1 It is a schematic diagram of the flow of the contour extraction method in the present invention. DETAILED DESCRIPTION
[0046] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention; it is obvious that the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments, and all other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making creative work are within the scope of protection of the present invention.
[0047] Example:
[0048] See also Figure 1 , a contour extraction method based on point cloud data, the specific implementation steps of this contour extraction method are as follows:
[0049] Step 1: Obtain point cloud data. At this time, data collection is performed using a 3D laser scanner to obtain point cloud data.
[0050] Then, each point of the target point cloud data is traversed by the traversal method;
[0051] Then, the RANSAC method is used to fit the local plane of the k-neighborhood of the data point coordinates. The specific implementation steps of the RANSAC method are as follows:
[0052] Step S1.1, randomly select a subset from the original data set as the hypothesized inliers;
[0053] Step S1.2, estimating model parameters based on the subset;
[0054] Step S1.3, traverse all data in the data set except the subset, if the data point is within the given error, mark it as an inlier, otherwise mark it as an outlier;
[0055] Step S1.4, all the inliers form a consistent set. If the number of points in the consistent set meets a given threshold, all the inliers in the consistent set are used to re-estimate the model parameters, and then the algorithm ends;
[0056] Step S1.5: If the number of inliers in the consistent set is less than the threshold, a new subset is selected and steps S1.1 to S1.4 are repeated;
[0057] Step S1.6: After iterations, select a consistent set with the largest number of inliers, use all the inliers in the consistent set to re-estimate the model parameters, and then terminate the algorithm.
[0058] In this process, a k-neighborhood search is performed in the local plane of the k-neighborhood to identify and remove noise points, and the k-neighborhood of the data point is analyzed to determine whether it is a noise point and perform corresponding processing.
[0059] Step 2: The traversed point cloud data set is projected onto the fitted local plane as the original data points;
[0060] When completing the projection of point cloud data to a local plane, the specific steps are as follows:
[0061] Step S2.1, loading point cloud data from a file;
[0062] Step S2.2: Prepare the camera parameters to describe the position and direction of the camera in three-dimensional space. The camera's intrinsic parameter matrix K is a 3×3 matrix, and the camera's extrinsic parameter matrix is:
[0063] T=[R,t]
[0064] Among them, R is a special orthogonal matrix, t is a three-dimensional vector;
[0065] Step S2.3, perform projection calculation, normalize the calculated coordinates to obtain pixel coordinates;
[0066] For each point p in the point cloud, the coordinates in the camera coordinate system are calculated as follows:
[0067]
[0068] in, are the homogeneous coordinates of the point p.
[0069] Step 3: Use the Gaussian kernel to eliminate the influence of noise points, perform noise reduction processing, and form a new local plane;
[0070] When the Gaussian kernel eliminates noise points, the mathematical model formula for calculation is:
[0071]
[0072] Where x and x′ are data points, is the standard deviation, and ||xx′|| 2 is the square of the Euclidean distance.
[0073] Step 4: Calculate the probability that each point is an edge point, and calculate the average of all probability groups;
[0074] Step 5: According to the average value of the probability group, the points with values less than the average value are boundary points, and those with values greater than the average value are internal points;
[0075] Step 6: According to the distribution of the determined boundary points and internal points, the result is obtained. The set of all boundary points constitutes the contour points of the target point cloud.
[0076] The above is only a preferred specific implementation of the present invention; however, the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical solution and its improved conception within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.
Claims
1. A contour extraction method based on point cloud data, characterized in that: The specific implementation steps of this contour extraction method are as follows: Step 1: Get the point cloud data, traverse each point of the target point cloud data through the traversal method, and then use the RANSAC method to fit the local plane of the k-neighborhood of the data point coordinates; Step 2: The traversed point cloud data set is projected onto the fitted local plane as the original data points; Step 3: Use the Gaussian kernel to eliminate the influence of noise points, perform noise reduction processing, and form a new local plane; Step 4: Calculate the probability that each point is an edge point, and calculate the average of all probability groups; Step 5: According to the average value of the probability group, the points with values less than the average value are boundary points, and those with values greater than the average value are internal points; Step 6: According to the distribution of the determined boundary points and internal points, the result is obtained. The set of all boundary points constitutes the contour points of the target point cloud.
2. The contour extraction method based on point cloud data according to claim 1, characterized in that: In the step 1, when acquiring point cloud data, data collection is performed by using a three-dimensional laser scanner to acquire point cloud data.
3. The contour extraction method based on point cloud data according to claim 1, characterized in that: In step 1, the specific implementation steps of the RANSAC method are as follows: Step S1.1, randomly select a subset from the original data set as the hypothesized inliers; Step S1.2, estimating model parameters based on the subset; Step S1.3, traverse all data in the data set except the subset, if the data point is within the given error, mark it as an inlier, otherwise mark it as an outlier; Step S1.4, all the inliers form a consistent set. If the number of points in the consistent set meets a given threshold, all the inliers in the consistent set are used to re-estimate the model parameters, and then the algorithm ends; Step S1.5: If the number of inliers in the consistent set is less than the threshold, a new subset is selected and steps S1.1 to S1.4 are repeated; Step S1.6: After iterations, select a consistent set with the largest number of inliers, use all the inliers in the consistent set to re-estimate the model parameters, and then terminate the algorithm.
4. The contour extraction method based on point cloud data according to claim 1, characterized in that: In the step 1, a k-neighborhood search is performed in the local plane of the k-neighborhood to identify and remove noise points, and the k-neighborhood of the data point is analyzed to determine whether it is a noise point and perform corresponding processing.
5. The contour extraction method based on point cloud data according to claim 1, characterized in that: In the step 2, the projection of the point cloud data to the local plane is completed, and the specific steps are as follows: Step S2.1, loading point cloud data from a file; Step S2.2, prepare camera parameters to describe the position and direction of the camera in three-dimensional space; Step S2.3: perform projection calculation, normalize the calculated coordinates to obtain pixel coordinates.
6. The contour extraction method based on point cloud data according to claim 5, characterized in that: In step S2.2, the camera's intrinsic parameter matrix K is a 3×3 matrix, and the camera's extrinsic parameter matrix is: T=[R,t] Where R is a special orthogonal matrix and t is a three-dimensional vector.
7. The contour extraction method based on point cloud data according to claim 5, characterized in that: In step S2.3, for each point p in the point cloud, the calculation formula of its coordinates in the camera coordinate system is: in, are the homogeneous coordinates of the point p.
8. The contour extraction method based on point cloud data according to claim 1, characterized in that: In step 3, when the Gaussian kernel eliminates noise points, the mathematical model formula for calculation is: Where x and x′ are data points, is the standard deviation, and ||xx′|| 2 is the square of the Euclidean distance.
Citation Information
Cited By
Irregular skin point cloud contour extraction method for micro-plane intersection projection
CN121169949A
