Point cloud salient feature detection method and device, and computer readable storage medium
By calculating the poles in the point cloud and performing median filtering and Gaussian difference, adaptive three-dimensional point cloud distinctive feature detection is realized, solving the problems of artificial detection's fatigue and high error detection rate, and improving detection accuracy and process control quality.
Patent Information
- Application Number
- CN202411357444.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-26
- Publication Date
- 2025-01-10
AI Technical Summary
In the prior art, the detection of distinctive feature of three-dimensional point clouds by manual inspection has fatigue, high error rate and large subjective judgment errors, which is not suitable for long-term quality process control.
By calculating the poles of the sampling points in the point cloud, dividing them into inner and outer poles, median filtering is performed to obtain local features of discrete surfaces, combined with Gaussian differences, a feature measure of global visual saliency is obtained to achieve adaptive Point cloud significant feature detection.
It effectively saves detection time, improves detection accuracy, enhances process control quality, reduces noise interference, and improves the accuracy of sharp edge features.
Smart Images
Figure CN119295729A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of image processing technology, and in particular to a point cloud significant feature detection method, device and computer-readable storage medium. Background Art
[0002] In the prior art, a three-dimensional point cloud is raw data obtained by discretely sampling the surface of an object through 3D digital scanning and other equipment, which contains the three-dimensional information of the object and related appearance attributes. The amount of point cloud data collected is generally huge, and the geometric characteristics of the object can be well reflected after three-dimensional reconstruction of the point cloud data. With the rapid development of technology, 3D point cloud processing methods have also been proposed for human-centric visual computing applications, such as object detection, recognition, positioning, and configuration based on 3D point clouds. The traditional curvature-based mathematical measurement method does not take into account the characteristics of the human visual system on the 3D model, and for the 3D model itself, the large amount of data is easily affected by noise and distorted.
[0003] At present, the extraction of significant features from geometric features actually predicts the human visual gaze point and eye movement, and screens out the areas of interest to the human eye, so that people can focus more on the significant feature areas when paying attention to the three-dimensional data model. In the field of 3D visual detection, saliency detection of 3D point clouds is used to find perceptually important areas that are unique to their surrounding areas and can be used to control the details of the process and accelerate the detection and classification of 3D objects.
[0004] However, manual visual inspection is inherently fatigue-prone, prone to false detections, and subject to subjective errors, making it unsuitable for long-term quality process control. Summary of the invention
[0005] In order to overcome the deficiencies in the prior art, the purpose of the present invention is to provide a point cloud significant feature detection method, device and computer-readable storage medium to solve the problems that current manual visual inspection is inherently fatigue-prone, prone to false detection, and has subjective judgment errors, and is not suitable for long-term quality process control.
[0006] The present invention proposes a method for detecting significant features of point clouds, which comprises:
[0007] Calculate the poles of the sampling points in the point cloud and divide the poles into inner poles and outer poles;
[0008] Perform median filtering on the distances from each sampling point to the corresponding inner pole and outer pole to obtain local features of the discrete surface;
[0009] The Gaussian differences of the local features of the discrete surface at different scales are superimposed to obtain a feature measure of global visual saliency.
[0010] Optionally, the calculating the poles of the sampling points in the point cloud specifically includes:
[0011] The preset sampling point set S = {s1, s2...sn}, for any point s in the sampling point set S i , point s i The corresponding Voronoi cell is Distance s in space i The space part closer to other points is recorded as the cell body V(s i );
[0012] The cell bodies V(s i ) constitutes the The intersection points of the partitions of the space are referred to as the poles, which are denoted as a set V(S).
[0013] Optionally, dividing the poles into inner poles and outer poles specifically includes:
[0014] For any point s in the sampling point set S i , and the set V(S) with s i The point with the greatest distance is taken as the inner pole;
[0015] The cell body V(s i ) relative to point s i The point on the other side that is farthest away is taken as the outer pole.
[0016] Optionally, the distances from each sampling point to the corresponding inner pole and outer pole are respectively subjected to median filtering to obtain the local features of the discrete surface, specifically including:
[0017] For any point s in the sampling point set S i , with point s i Construct a sphere b as the center r (s i ), the set of sampling points in the sphere is
[0018] For the sampling point set Each sampling point s in i The distance to the corresponding inner pole and outer pole is respectively subjected to median filtering, and λmax and λmin are respectively the larger and smaller values of the results obtained after filtering, then the sampling point s i The characteristic measure of is defined as F(s i )=-log(λmax / λmin).
[0019] Optionally, the superimposing Gaussian differences of the local features of the discrete surface at different scales to obtain a feature measure of global visual saliency specifically includes:
[0020] The Gaussian pyramid model is used to measure multiple scales and calculate each sampling point s i The prominence of
[0021] Overlay each sampling point s i The saliency feature of the global visual saliency is obtained.
[0022] Optionally, the method further comprises:
[0023] Create preset feature measurement algorithms;
[0024] The sampling point s is calculated by the feature measurement algorithm i The characteristic measure F(s i ).
[0025] Optionally, the creating a preset feature measurement algorithm specifically includes:
[0026] Taking the sampling point set S and the preset cube vertex set B as algorithm input;
[0027] Each sampling point s in the point cloud i The characteristic measure F(s i ) as the algorithm output;
[0028] The point cloud is placed into the cube, and the vertex set of the cube is recorded as set B.
[0029] Optionally, the step of creating a preset feature measurement algorithm further includes:
[0030] The Voronoi diagram of S∪B is calculated to obtain the pole set P; wherein the inner pole is denoted as The outer pole is denoted by The set of extreme points is denoted by P = P + ∪P - .
[0031] The present invention also proposes a point cloud significant feature detection device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the point cloud significant feature detection method as described in any one of the above items.
[0032] The present invention also proposes a computer-readable storage medium, on which a point cloud significant feature detection program is stored. When the point cloud significant feature detection program is executed by a processor, the steps of the point cloud significant feature detection method as described in any of the above items are implemented.
[0033] The method, device and computer-readable storage medium for detecting significant features of point clouds of the present invention are implemented by calculating the poles of the sampling points in the point cloud and dividing the poles into inner poles and outer poles; median filtering is performed on the distances from each sampling point to the corresponding inner poles and outer poles to obtain local features of discrete surfaces; Gaussian differences of local features of discrete surfaces at different scales are superimposed to obtain feature measures of global visual saliency. An adaptive point cloud significant feature detection scheme is implemented, which effectively saves detection time, improves detection accuracy and enhances process control quality for three-dimensional reconstruction models. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:
[0035] Figure 1 is a first flow chart of the method for detecting significant features of point clouds according to the present invention;
[0036] Figure 2 is a second flow chart of the method for detecting significant features of point clouds according to the present invention;
[0037] Figure 3 is a third flow chart of the method for detecting significant features of point clouds of the present invention;
[0038] Figure 4 is a fourth flow chart of the method for detecting significant features of point clouds of the present invention;
[0039] Figure 5 is a fifth flow chart of the method for detecting significant features of point clouds of the present invention;
[0040] Figure 6 is a sixth flow chart of the method for detecting significant features of point clouds of the present invention;
[0041] Figure 7 Schematic diagram of the extreme points of the point cloud salient feature detection method of the present invention. DETAILED DESCRIPTION
[0042] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.
[0043] In the subsequent description, the suffixes such as "module", "component" or "unit" used to represent elements are only used to facilitate the description of the present invention, and have no specific meanings. Therefore, "module", "component" or "unit" can be used in a mixed manner.
[0044] Figure 1 This is the first flow chart of the method for detecting significant features of point clouds of the present invention. This embodiment proposes a method for detecting significant features of point clouds, the method comprising:
[0045] S1. Calculate the poles of the sampling points in the point cloud, and divide the poles into inner poles and outer poles;
[0046] S2, performing median filtering on the distances from each sampling point to the corresponding inner pole and outer pole to obtain local features of the discrete surface;
[0047] S3. Superimposing Gaussian differences of local features of the discrete surface at different scales to obtain a feature measure of global visual saliency.
[0048] In this embodiment, a point cloud salient feature detection scheme based on Voronoi poles is proposed. First, the local features of the discrete surface are calculated through the relationship between the Voronoi inner and outer poles and the sampling points. Then, the Gaussian difference of local features at different scales is superimposed to obtain the feature measure of global visual saliency.
[0049] For details, please refer to Figure 2 The second flow chart of the method for detecting significant features of point clouds of the present invention is shown. In which, step 1, 3D point cloud data is imported, step 2, the Voronoi poles of each sampling point are calculated, step 3, the Voronoi poles of the sampling points are distinguished, that is, they are divided into two types: internal poles and external poles, step 4, the distances from each sampling point to its own internal pole and external pole are calculated, so as to obtain the significant feature measurement value of each sampling point, step 5, multi-scale measurement using the Gaussian pyramid model, step 6, superimposing features to obtain the final significant feature value.
[0050] It can be seen that this embodiment, by implementing the above-mentioned Voronoi pole-based point cloud significant feature calculation, can effectively detect and measure the visual significance of point cloud data points, and effectively avoid the problems of long time consumption, high misjudgment rate and high false detection rate caused by manual detection of three-dimensional reconstruction models.
[0051] The beneficial effect of this embodiment is that by calculating the poles of the sampling points in the point cloud and dividing the poles into inner poles and outer poles; median filtering the distances from each sampling point to its corresponding inner pole and outer pole, the local features of the discrete surface are obtained; and the Gaussian difference of the local features of the discrete surface at different scales is superimposed to obtain the feature measure of global visual saliency. An adaptive point cloud salient feature detection scheme is implemented, which effectively saves detection time, improves detection accuracy, and enhances process control quality for three-dimensional reconstruction models.
[0052] Figure 3 is a third flow chart of the method for detecting significant features of point clouds of the present invention. In this embodiment, the step of calculating the poles of the sampling points in the point cloud specifically includes:
[0053] S11, preset sampling point set S = {s1, s2...sn}, for any point s in the sampling point set S, i , point s i The corresponding Voronoi cell is Distance s in space i The space part closer to other points is recorded as the cell body V(s i );
[0054] S12, the cell bodies V (s i ) constitutes the The intersection points of the partitions of the space are referred to as the poles, which are denoted as a set V(S).
[0055] Figure 7 is a schematic diagram of the extreme points of the method for detecting significant features of point clouds of the present invention. In this embodiment, the space D=2 is used as an example. Please refer to Figure 7 , the black dots are sampling points, and the intersection of the straight lines are the Voronoi poles.
[0056] Figure 4 is a fourth flow chart of the method for detecting significant features of point clouds of the present invention. In this embodiment, the step of dividing the poles into inner poles and outer poles specifically includes:
[0057] S13. For any point s in the sampling point set S, i , and the set V(S) with s i The point with the greatest distance is taken as the inner pole;
[0058] S14, the cell body V (s i ) relative to point s i The point on the other side that is farthest away is taken as the outer pole.
[0059] In this embodiment, for each point s in the sampling point set S i , which has two extreme points, namely i The farthest point and the cell body V(s i ) relative to s i The points on the other side that are farthest away are called the inner pole and outer pole respectively.
[0060] Optionally, in this embodiment, the poles are divided into poles inside the object and poles outside the object through a preset Sculpting algorithm.
[0061] Figure 5 is a fifth flow chart of the method for detecting significant features of point clouds of the present invention. In this embodiment, the distances from each sampling point to its corresponding inner pole and outer pole are respectively subjected to median filtering to obtain the steps of obtaining local features of discrete surfaces, specifically including:
[0062] S21. For any point s in the sampling point set S, i , with point s i Construct a sphere b as the center r (s i ), the set of sampling points in the sphere is
[0063] S22, the sampling point set Each sampling point s in i The distance to the corresponding inner pole and outer pole is respectively subjected to median filtering, and λmax and λmin are respectively the larger and smaller values of the results obtained after filtering, then the sampling point s i The characteristic measure of is defined as F(s i )=-log(λmax / λmin).
[0064] In this embodiment, a preset feature measurement algorithm is created; and the sampling point s is calculated by the feature measurement algorithm. i The characteristic measure F(s i ).
[0065] Specifically, on the one hand, the sampling point set S and the preset cube vertex set B are used as algorithm inputs; on the other hand, each sampling point s in the point cloud is i The characteristic measure F(s i ) as the algorithm output; wherein, the point cloud is placed into the cube, and the vertex set of the cube is recorded as set B.
[0066] In this embodiment, the detection method further includes: performing Voronoi diagram calculation on S∪B to obtain a pole set P; wherein the inner pole is recorded as The outer pole is denoted by The set of extreme points is denoted by P = P + ∪P - .
[0067] Figure 66 is a flow chart of the method for detecting salient features of point clouds of the present invention. In this embodiment, the step of superimposing Gaussian differences of local features of the discrete surface at different scales to obtain feature measures of global visual saliency specifically includes:
[0068] S31, using Gaussian pyramid model multi-scale measurement, calculate each sampling point s i The prominence of
[0069] S32, superimpose each sampling point s i The saliency feature of the global visual saliency is obtained.
[0070] In this embodiment, based on the above detection scheme, the Voronoi poles can be used to approximate the central axis of the geometric body, and the central axis can reflect the basic contour information of the object. Therefore, this embodiment uses the distance from the pole to the sampling point to measure the local characteristics.
[0071] In this embodiment, on the one hand, for areas with large curvature, the Voronoi sphere with the Voronoi vertex as the center has a smaller radius, that is, in areas with large (small) curvature, the Voronoi cell vertex corresponding to the sampling point is closer (farther) from the surface of the object; on the other hand, for a certain sampling point, since the direction of its curvature may be toward the inside or outside of the object, the radii of the poles corresponding to the inside and outside are quite different. Therefore, for points that bend outward, the poles outside may be at infinity. The above scheme of this embodiment mainly measures the local features of the point cloud based on the ratio of the distances from the inner and outer poles to the sampling points. Therefore, the significant feature measurement method involved in the above scheme of this embodiment has better anti-interference performance in the presence of noise than the traditional local feature method based on curvature, and has higher accuracy for the feature measurement of sharp edges.
[0072] Based on the above embodiments, the present invention also proposes a point cloud significant feature detection device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the computer program is executed by the processor, the steps of the point cloud significant feature detection method as described in any one of the above items are implemented.
[0073] It should be noted that the above-mentioned device embodiment and method embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, and the technical features in the method embodiment are correspondingly applicable in the device embodiment, which will not be repeated here.
[0074] Based on the above embodiments, the present invention also proposes a computer-readable storage medium, which stores a point cloud significant feature detection program. When the point cloud significant feature detection program is executed by a processor, the steps of the point cloud significant feature detection method as described in any one of the above items are implemented.
[0075] It should be noted that the above-mentioned medium embodiment and method embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, and the technical features in the method embodiment are correspondingly applicable in the medium embodiment, which will not be repeated here.
[0076] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present invention.
[0077] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.
[0078] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.
Claims
1. A method for detecting significant features of a point cloud, characterized in that: The method comprises: Calculate the poles of the sampling points in the point cloud and divide the poles into inner poles and outer poles; Perform median filtering on the distances from each sampling point to the corresponding inner pole and outer pole to obtain local features of the discrete surface; The Gaussian differences of the local features of the discrete surface at different scales are superimposed to obtain a feature measure of global visual saliency.
2. The method for detecting significant features of point clouds according to claim 1, characterized in that: The calculation of the extreme points of the sampling points in the point cloud specifically includes: The preset sampling point set S = {s1, s2...sn}, for any point s in the sampling point set S i , point s i The corresponding Voronoi cell is Distance s in space i The space part closer to other points is recorded as the cell body V(s i ); The cell bodies V(s i ) constitutes the The intersection points of the partitions of the space are referred to as the poles, which are denoted as a set V(S).
3. The method for detecting significant features of point cloud according to claim 2, characterized in that: The dividing the poles into inner poles and outer poles specifically includes: For any point s in the sampling point set S i , and the set V(S) with s i The point with the greatest distance is taken as the inner pole; The cell body V(s i ) relative to point s i The point on the other side that is farthest away is taken as the outer pole.
4. The method for detecting significant features of point cloud according to claim 3, characterized in that: The distances from each sampling point to the corresponding inner pole and outer pole are respectively subjected to median filtering to obtain local features of the discrete surface, specifically including: For any point s in the sampling point set S i , with point s i Construct a sphere b as the center r (s i ), the set of sampling points in the sphere is For the sampling point set Each sampling point s in i The distance to the corresponding inner pole and outer pole is respectively subjected to median filtering, and λmax and λmin are respectively the larger and smaller values of the results obtained after filtering, then the sampling point s i The characteristic measure of is defined as F(s i )=-log(θmax / θmin).
5. The method for detecting significant features of point cloud according to claim 4, characterized in that: The Gaussian difference of the local features of the discrete surface at different scales is superimposed to obtain a feature measure of global visual saliency, specifically including: The Gaussian pyramid model is used to measure multiple scales and calculate each sampling point s i The prominence of Overlay each sampling point s i The saliency feature of the global visual saliency is obtained.
6. The method for detecting significant features of point clouds according to claim 4, characterized in that: The method further comprises: Create preset feature measurement algorithms; The sampling point s is calculated by the feature measurement algorithm i The characteristic measure F(s i ).
7. The method for detecting significant features of point cloud according to claim 6, characterized in that: The step of creating a preset feature measurement algorithm specifically includes: Taking the sampling point set S and the preset cube vertex set B as algorithm input; Each sampling point s in the point cloud i The characteristic measure F(s i ) as the algorithm output; The point cloud is placed into the cube, and the vertex set of the cube is recorded as set B.
8. The method for detecting significant features of point clouds according to claim 6, characterized in that: The step of creating a preset feature measurement algorithm further includes: The Voronoi diagram of S∪B is calculated to obtain the pole set P; wherein the inner pole is denoted as The outer pole is denoted by The set of extreme points is denoted by P = P + ∪P - .
9. A point cloud significant feature detection device, characterized in that: The device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the steps of the point cloud significant feature detection method according to any one of claims 1 to 8 are implemented.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a point cloud significant feature detection program, which, when executed by a processor, implements the steps of the point cloud significant feature detection method according to any one of claims 1 to 8.