CN109816664B - Three-dimensional point cloud segmentation method and device - Google Patents
Three-dimensional point cloud segmentation method and device Download PDFInfo
- Publication number
- CN109816664B CN109816664B CN201811589435.9A CN201811589435A CN109816664B CN 109816664 B CN109816664 B CN 109816664B CN 201811589435 A CN201811589435 A CN 201811589435A CN 109816664 B CN109816664 B CN 109816664B
- Authority
- CN
- China
- Prior art keywords
- point cloud
- dimensional point
- dimensional
- area
- points
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Image Processing (AREA)
Abstract
The invention discloses a three-dimensional point cloud segmentation method and a device, which are used for acquiring three-dimensional point clouds in a direction parallel to a plane area; preprocessing the three-dimensional point cloud to obtain a preprocessed three-dimensional point cloud; according to the depth value of each three-dimensional point in the preprocessed three-dimensional point cloud, obtaining a three-dimensional point cloud of a non-planar area, and obtaining an initial three-dimensional point cloud of a concave area; processing all three-dimensional points in the three-dimensional point cloud of the initial concave area by adopting an edge searching method to obtain a plurality of three-dimensional points to form the edge of the concave area; obtaining a three-dimensional point cloud of the concave area; according to the segmentation method provided by the invention, errors and redundant point clouds are removed by filtering and downsampling methods, the number of three-dimensional points in the three-dimensional point clouds is reduced, and the segmentation speed of pit and groove areas is effectively improved; the depth threshold is utilized to remove the plane area, the non-plane area is reserved, the approximate area of the pit can be rapidly segmented, the complexity of the segmentation method is reduced, and the segmentation speed is improved.
Description
Technical Field
The invention relates to a three-dimensional point cloud processing method and device, in particular to a three-dimensional point cloud segmentation method and device.
Background
The three-dimensional point cloud is the most common and basic three-dimensional model in a spatial information expression mode. The point cloud model is often obtained directly by measurement, each point corresponds to a measurement point and is not processed by other methods, so that the point cloud model contains the maximum information quantity of a target. The three-dimensional point cloud is mainly used for representing data in a three-dimensional space, and compared with a planar two-dimensional graph, the three-dimensional point cloud has depth direction information, and can naturally decouple a target and a background in the depth direction. In addition, for three-dimensional vision measurement, the three-dimensional point cloud is a main supporting means, and the actual physical geometric information of the target in space can be obtained through extracting the three-dimensional point cloud of the target.
In practical application, because the actually obtained point cloud contains a large amount of noise, the obtained three-dimensional point cloud does not contain the complete information of the target, and rotation transformation exists at different angles, so that the shape of the same target point cloud is quite different, especially when facing the three-dimensional point cloud of a pavement containing both a concave area and a plane area, such as a three-dimensional point cloud of a pavement containing pits, the pavement is a plane, but the concave area of the pits is included in the plane, and when the three-dimensional point cloud is segmented again, the segmentation difficulty is high because the segmentation line between the concave area and the plane area cannot be found accurately due to more noise, so that the three-dimensional point cloud segmentation method in the prior art cannot segment accurately when facing the three-dimensional point cloud containing the concave area.
Disclosure of Invention
The invention aims to provide a three-dimensional point cloud segmentation method and a three-dimensional point cloud segmentation device, which are used for solving the problem that the three-dimensional point cloud segmentation method in the prior art cannot accurately segment when facing the three-dimensional point cloud comprising a concave area.
In order to realize the tasks, the invention adopts the following technical scheme:
a three-dimensional point cloud segmentation method for segmenting a three-dimensional point cloud of a concave region from a three-dimensional point cloud containing the concave region and a plane region, the method comprising the steps of:
step 1, collecting three-dimensional point clouds containing a concave area and a plane area in a direction parallel to the plane area;
step 2, preprocessing the three-dimensional point cloud to obtain a preprocessed three-dimensional point cloud, wherein the preprocessing comprises a downsampling operation and a filtering operation which are sequentially executed;
step 3, deleting all three-dimensional points with depth values smaller than a depth threshold value according to the depth value of each three-dimensional point in the preprocessed three-dimensional point cloud to obtain a three-dimensional point cloud of a non-planar area, wherein the depth threshold value is twice the average value of the depth values of all three-dimensional points in the preprocessed three-dimensional point cloud;
step 4, taking the three-dimensional point with the largest depth value as a seed point in the three-dimensional point cloud of the non-planar area, and obtaining an initial three-dimensional point cloud of the concave area by adopting an area growth method;
step 5, processing all three-dimensional points in the three-dimensional point cloud of the initial concave area by adopting an edge searching method to obtain a plurality of three-dimensional points to form a concave area edge;
and 6, collecting all three-dimensional points within the edge of the concave area to obtain the concave area three-dimensional point cloud.
Further, the step 2 is performed according to the following steps:
step 21, downsampling the three-dimensional point cloud by using a voxel grid filtering method to obtain the downsampled three-dimensional point cloud;
step 22, performing direct filtering on the down-sampled three-dimensional point cloud to obtain a direct filtered three-dimensional point cloud;
and step 23, carrying out statistical filtering on the three-dimensional point cloud after the direct filtering to obtain the processed three-dimensional point cloud.
Further, in the step 3, according to the depth value of each three-dimensional point in the preprocessed three-dimensional point cloud, deleting all three-dimensional points with depth values smaller than the depth threshold value by adopting a RANSAC algorithm, and obtaining the three-dimensional point cloud of the non-planar area.
Further, when the edge searching method is used for processing all three-dimensional points in the three-dimensional point cloud of the initial concave area, one vertex of the three-dimensional point cloud of the initial concave area is used as an initial boundary point, and a plurality of three-dimensional points are sequentially obtained to form the edge of the concave area by utilizing the principle of the maximum angle of the boundary point.
The device comprises a three-dimensional point cloud acquisition module, a three-dimensional point cloud preprocessing module, a planar area three-dimensional point cloud removal module and a recessed area three-dimensional point cloud segmentation module;
the three-dimensional point cloud acquisition module is connected with the three-dimensional point cloud preprocessing module and is used for acquiring the three-dimensional point cloud containing the concave area and the plane area in a direction parallel to the plane area;
the three-dimensional point cloud preprocessing module is connected with the plane area three-dimensional point cloud removing module and is used for preprocessing the three-dimensional point cloud to obtain a preprocessed three-dimensional point cloud, and the preprocessing operation comprises a downsampling operation and a filtering operation which are sequentially executed;
the planar area three-dimensional point cloud removing module is connected with the concave area three-dimensional point cloud dividing module and is used for taking a three-dimensional point with the largest depth value as a seed point in the three-dimensional point cloud of the non-planar area and obtaining an initial concave area three-dimensional point cloud by adopting an area growth method;
the method is also used for processing all three-dimensional points in the three-dimensional point cloud of the initial concave area by adopting an edge searching method to obtain a plurality of three-dimensional points to form the concave area edge;
and the three-dimensional point cloud of the concave area is obtained by collecting all three-dimensional points within the edge of the concave area.
Further, the three-dimensional point cloud preprocessing module, the planar area three-dimensional point cloud removing module and the concave area three-dimensional point cloud segmentation module are all intelligent mobile devices.
Further, the intelligent mobile device is a mobile phone.
Compared with the prior art, the invention has the following technical effects:
1. when the three-dimensional point cloud segmentation method and device provided by the invention are used for segmenting the concave region, errors and redundant point clouds are removed by filtering, downsampling and other methods, the number of three-dimensional points in the three-dimensional point cloud is reduced, and the segmentation speed of the concave region is effectively improved;
2. according to the three-dimensional point cloud segmentation method and device, when the concave region is segmented, the depth threshold is utilized to remove the plane region, the non-plane region is reserved, the rough region of the concave region can be rapidly segmented, the complexity of the segmentation method is reduced, and the segmentation speed is improved;
3. when the three-dimensional point cloud segmentation method and the three-dimensional point cloud segmentation device provided by the invention segment the concave region, an initial concave region is obtained by adopting a region generation method, and an edge searching method is used on the basis of the initial concave region to obtain the edge of the complete concave region, so that the integrity and the accuracy of segmentation of the concave region are improved;
4. according to the three-dimensional point cloud segmentation method and device, when the concave region is segmented, the boundary point maximum angle principle is utilized to sequentially obtain a plurality of three-dimensional points to form the edge of the concave region, so that the accuracy of segmentation of the concave region is further improved.
Drawings
FIG. 1 is a flow chart of a three-dimensional point cloud segmentation method provided by the invention;
FIG. 2 is a left and right camera image when used to reconstruct a three-dimensional point cloud of a road surface in accordance with one embodiment of the present invention;
FIG. 3 is a three-dimensional point cloud of a road surface acquired in one embodiment of the invention;
FIG. 4 is a three-dimensional point cloud of a pothole area obtained by segmentation in one embodiment of the present invention;
fig. 5 is a schematic structural diagram of a three-dimensional point cloud segmentation apparatus according to an embodiment of the present invention.
Detailed Description
Three-dimensional point cloud: each three-dimensional point has an abscissa value, an ordinate value and a depth value, and the left lens of the camera is used as the origin of the coordinate system, and the abscissa value, the ordinate value and the depth value are generally positive values.
Boundary point maximum angle principle: among the plurality of adjacent points of a certain boundary point, the boundary point and any adjacent point have included angles (the included angle does not include other adjacent points), so that a plurality of included angles exist, and two adjacent points corresponding to the difference value between the two largest included angles are necessarily boundary points.
Example 1
A three-dimensional point cloud segmentation method for segmenting a three-dimensional point cloud of a concave region from a three-dimensional point cloud containing the concave region and a plane region, as shown in fig. 1, the segmentation method is performed according to the following steps:
step 1, collecting three-dimensional point clouds containing a concave area and a plane area in a direction parallel to the plane area;
in this step, first, a three-dimensional point cloud including a concave area and a planar area is acquired by using a three-dimensional imaging device such as a binocular stereo camera, and the three-dimensional point cloud is used as an input of the measurement method provided in this embodiment.
The imaging device is noted to be opposite to plane shooting during shooting, and the whole concave area range is covered.
In this embodiment, the road surface three-dimensional point cloud obtained by three-dimensional reconstruction of the left and right camera images of the binocular camera shown in fig. 2 is shown in fig. 3.
In the step, when the three-dimensional point cloud is segmented, the existing methods comprise an Euclidean algorithm, a kdToe algorithm and an OcTree algorithm, but the problems of low segmentation accuracy rate of the above algorithms can occur when the three-dimensional point cloud of the road surface faces.
Because the acquired three-dimensional point cloud contains a large amount of data and contains noise, if the three-dimensional point cloud is directly processed, the phenomena of low segmentation speed and poor segmentation effect can occur, so in the step, the three-dimensional point cloud is firstly subjected to dimension reduction and then is filtered.
Step 2, preprocessing the three-dimensional point cloud to obtain a preprocessed three-dimensional point cloud, wherein the preprocessing comprises a downsampling operation and a filtering operation which are sequentially executed;
firstly, the three-dimensional point cloud is downsampled, the voxel grid filtering method is adopted to downsample the three-dimensional point cloud, so that redundant data quantity is reduced, the point cloud processing speed is improved, details can be kept in the depth direction during filtering, and detail loss in the depth direction is reduced.
The specific process is as follows: creating a three-dimensional voxel grid through the input three-dimensional point cloud, wherein the voxel grid is equivalent to a set of a space three-dimensional cube, and then in each voxel, using the centers of gravity of all points in the voxel to approximate and display other points in the voxel, so that all points in the voxel grid are finally represented by a center of gravity point, and all point clouds are processed according to the voxel grid to obtain down-sampling point clouds, so that accurate representation of a curved surface corresponding to the sampling point is obtained.
Step 22, performing direct filtering on the down-sampled three-dimensional point cloud to obtain a direct filtered three-dimensional point cloud;
in this step, the downsampled three-dimensional point cloud is filtered, and a direct filtering method is adopted, specifically, a depth direction data range of the three-dimensional point cloud is set, for example, between 0.8 and 1.2m, and point clouds outside the range are all removed as interference points.
And step 23, carrying out statistical filtering on the three-dimensional point cloud after the direct filtering to obtain the processed three-dimensional point cloud.
In the step, a statistical filtering method is adopted for the three-dimensional point cloud after the direct filtering, a statistical analysis is carried out on the neighborhood of each point, some point clouds which do not meet the standard are removed, isolated redundant point clouds are reduced, a statistical analysis is carried out on the neighborhood of each point through the statistical filtering, and the points which do not meet the set standard are trimmed. Specifically, a sparse outlier removal method is employed in the present embodiment, based on the computation of the distance distribution from point to point in the input three-dimensional point cloud. For each three-dimensional point, the average distance from the three-dimensional point to all adjacent points is calculated, the result is a Gaussian distribution, the shape of the Gaussian distribution is determined by the mean and the standard deviation, and the three-dimensional points with the average distance outside the standard range (defined by the average value and the variance of the global distance) are outliers and are removed from the three-dimensional point cloud.
When the concave region is segmented in the embodiment, the depth threshold is utilized to remove the plane region, and the non-plane region is reserved, so that the concave region can be rapidly segmented roughly, the complexity of the segmentation method is reduced, and the segmentation speed is improved.
Step 3, deleting all three-dimensional points with depth values smaller than a depth threshold value according to the depth value of each three-dimensional point in the preprocessed three-dimensional point cloud to obtain a three-dimensional point cloud of a non-planar area, wherein the depth threshold value is twice the average value of the depth values of all three-dimensional points in the preprocessed three-dimensional point cloud;
in this step, since the three-dimensional point cloud is obtained parallel to the planar area at the time of obtaining, depth values in the three-dimensional point cloud of the planar area should be the same, and by this feature, the planar three-dimensional point cloud can be found.
Preferably, according to the depth value of each three-dimensional point in the preprocessed three-dimensional point cloud, deleting all three-dimensional points with depth values smaller than a depth threshold value by adopting a RANSAC algorithm to obtain the three-dimensional point cloud of the non-planar area.
In this step, when the RANSAC algorithm is used to extract the plane, a depth threshold is set, where the depth threshold is set according to the thickness of the plane, for example, when the pit area of the road surface is divided, a road plane with a depth of 2mm is to be divided, and in the measurement method provided in this embodiment, all points within a depth of 2mm near the road plane are taken as three-dimensional points of the road plane, which is equivalent to cutting the three-dimensional point cloud in the depth direction, and cutting the road plane.
In this step, the planar point cloud is extracted and the other non-planar partial point cloud (including the concave point cloud) has been segmented. Under normal conditions, due to noise and interference of point cloud points without matching errors, complete and accurate concave partial point cloud cannot be provided at the moment, and further processing is needed.
Therefore, optionally, before executing the step 23, the three-dimensional point cloud of the non-planar area is first subjected to a second filtering by adopting a statistical filtering manner, so as to obtain a three-dimensional point cloud of the non-planar area after filtering, and the three-dimensional point cloud of the non-planar area after filtering is input into the step 23.
In this step, a statistical filtering mode is adopted again during the secondary filtering, filtering parameters are set, isolated point clouds with a smaller useless range are removed, the total number of three-dimensional points contained in a certain distance of each point cloud is set, if the total number of three-dimensional points is smaller than a set minimum three-dimensional point total number threshold value, the isolated point clouds are considered to be the isolated point clouds, and if the total number of three-dimensional points is larger than a set maximum three-dimensional point total number threshold value, the isolated point clouds are considered to be the ineffective three-dimensional point clouds. For example, counting the number of three-dimensional points within 5 cm around a certain three-dimensional point, if the counted result is less than 500, regarding the three-dimensional point and all three-dimensional points within 5 cm around the three-dimensional point as isolated point clouds, if the counted result is more than 50000, regarding invalid point clouds exceeding the theoretical concave data amount, and removing, for example, in the pit area in fig. 3, small pit areas beside the largest pit area, and removing the small pit areas in this step, wherein the small pit areas are isolated point clouds.
Step 4, taking the three-dimensional point with the largest depth value as a seed point in the three-dimensional point cloud of the non-planar area, and obtaining an initial three-dimensional point cloud of the concave area by adopting an area growth method;
in this step, according to the depth threshold set during downsampling, which is used as the distance threshold in the region growing method, the three-dimensional point with the largest depth value is used as the starting point, the point within the range of the distance threshold is searched, the new starting point is iterated by the newly searched point, and the searching is continued until the three-dimensional point at the edge of the segmented concave region is searched, so as to obtain the initial three-dimensional point cloud of the concave region.
In order to improve the accuracy of segmentation, in this embodiment, the three-dimensional point cloud of the initial concave region is subjected to another edge search to obtain the edges of the complete concave region.
Step 5, processing all three-dimensional points in the three-dimensional point cloud of the initial concave area by adopting an edge searching method to obtain a plurality of three-dimensional points to form a concave area edge;
in this step, the method of edge search may be a depth threshold detection method, a gradient difference threshold detection method, or the like, but in the prior art, when facing a recessed region in a plane, particularly when the edge of the recess has a slope, the edge of the recessed region cannot be found well by the existing threshold detection method, resulting in inaccurate division of the recessed region.
Therefore, in this embodiment, an iterative edge search method is provided, which uses a vertex of the three-dimensional point cloud of the initial concave region as an initial boundary point, and sequentially obtains edges of the concave region formed by a plurality of three-dimensional points by using the principle of maximum angles of the boundary points.
Specifically, the method is carried out according to the following steps:
step I, clearing a boundary three-dimensional point set;
II, adopting a k adjacent algorithm to find a plurality of adjacent three-dimensional points around the initial boundary point;
in the step, the initial boundary point is taken as a sphere center, a distance threshold is taken as a sphere radius, a selection sphere is established, all three-dimensional points in the selection sphere are taken as adjacent three-dimensional points, a plurality of adjacent three-dimensional points are obtained, and the distance threshold is twice the average value of depth values of all three-dimensional points in the three-dimensional point cloud after secondary filtering;
step III, calculating the included angle between each adjacent three-dimensional point and the initial boundary point to obtain a plurality of included angle values;
step IV, arranging all the included angle values according to descending order of magnitude, and calculating the angle difference between two adjacent included angle values to obtain a plurality of angle differences;
step V, selecting the maximum angle difference from all the angle differences, and taking two adjacent three-dimensional points corresponding to two included angle values corresponding to the maximum angle difference as two boundary three-dimensional points;
step VI, if the boundary three-dimensional point set is empty, executing a step VIII; otherwise, executing VII;
step VII, judging whether one of the two boundary three-dimensional points obtained in the step V is in the boundary three-dimensional point set;
if one of the boundary three-dimensional point sets is in the boundary three-dimensional point set, taking another boundary three-dimensional point which is not in the boundary three-dimensional point set as a new initial boundary point, putting the new initial boundary point into the boundary three-dimensional point set, and executing the step IX;
if the two boundary three-dimensional points are in the boundary three-dimensional point set, selecting one point with the farthest distance from the initial three-dimensional point as a new initial boundary point;
otherwise, firstly putting the two boundary three-dimensional points into a boundary three-dimensional point set, respectively calculating the angle values between the two boundary three-dimensional points and the initial boundary point, and executing the step IX after selecting the boundary three-dimensional point with the largest angle value as a new initial boundary point;
step IX, taking a new initial boundary point as input, repeating the steps II to IX until the new initial three-dimensional point obtained at this time is one vertex of the initial concave area three-dimensional point cloud selected when the step II is executed for the first time, and executing the step X;
in the embodiment, the boundary point maximum angle principle is utilized to sequentially obtain a plurality of three-dimensional points to form the edge of the concave region, so that the accuracy of concave segmentation is further improved.
And step X, extracting all boundary three-dimensional points in the boundary three-dimensional point set from the initial concave region three-dimensional point cloud to obtain a plurality of three-dimensional points to form a concave region edge.
In the step X, after all the boundary three-dimensional points in the boundary three-dimensional point set are extracted from the initial three-dimensional point cloud of the concave area, if the three-dimensional point cloud of the concave area is obtained directly through the sparse boundary three-dimensional points, noise points exist in the three-dimensional point cloud of the concave area, that is, the situation of inaccurate segmentation exists, so that in the embodiment, all the boundary three-dimensional points are condensed to obtain a plurality of three-dimensional points to form the edge of the concave area.
In this step, the densification method may be interpolation or the like, but in this embodiment, in order to improve the accuracy of densification, the segmented concave region may be more complete and accurate, and the boundary densification is to evolve the sparse boundary point into a dense point cloud, or to add the boundary point on the sparse point line into the boundary point cloud.
Specifically, all boundary three-dimensional points are densified in this embodiment according to the following steps:
step A, finding two adjacent boundary three-dimensional points from all the boundary three-dimensional points to be respectively used as an endpoint and an endpoint;
b, taking the end point as the center of the sphere, taking the distance threshold value as the radius of the sphere, establishing a selection sphere, taking all three-dimensional points in the selection sphere as adjacent boundary three-dimensional points, and obtaining a plurality of adjacent boundary three-dimensional points;
the distance threshold in this step is the same as the distance threshold in step II.
Step C, calculating the distances from all the adjacent boundary three-dimensional points to the straight line formed by the end points and the end points, and obtaining distance values of a plurality of adjacent boundary three-dimensional points;
step D, using the adjacent boundary three-dimensional points corresponding to the distance values smaller than the dense distance threshold value as dense boundary three-dimensional points to obtain a plurality of dense boundary three-dimensional points, wherein the dense distance threshold value is the density value of the three-dimensional point cloud after secondary filtering;
step E, repeating the step A to the step E until all the boundary three-dimensional points are used as end points, and obtaining a plurality of dense boundary three-dimensional points;
and F, combining all dense boundary three-dimensional points with all boundary three-dimensional points to obtain a plurality of three-dimensional points to form a concave region edge.
And step 25, collecting all three-dimensional points within the edge of the concave area to obtain the concave area three-dimensional point cloud.
Thus, the edge of the concave region is the edge composed of the dense boundary points and all boundary three-dimensional points obtained in the step F, and all three-dimensional points within the edge are taken out, including the edge, to obtain a concave region three-dimensional point cloud.
In this embodiment, for the three-dimensional point cloud image of the road surface shown in fig. 3, the method of edge search and edge densification provided in this embodiment is adopted to obtain a pit area shown in fig. 4, and the edge of the pit area is composed of dense three-dimensional boundary points and three-dimensional boundary points.
When the concave region is segmented in the embodiment, an initial concave region is obtained by adopting a region generation method, and an edge searching method is used on the basis of the initial concave region to obtain the edge of the complete concave region, so that the segmentation integrity and accuracy of the concave region are improved, and a densification method of sparse edges is provided to obtain a more accurate concave region.
Example two
The embodiment discloses a three-dimensional point cloud segmentation device, which is used for the three-dimensional point cloud segmentation method in the first embodiment, as shown in fig. 5, wherein the device comprises a three-dimensional point cloud acquisition module, a three-dimensional point cloud preprocessing module, a planar area three-dimensional point cloud removal module and a recessed area three-dimensional point cloud segmentation module;
the three-dimensional point cloud acquisition module is connected with the three-dimensional point cloud preprocessing module and is used for acquiring the three-dimensional point cloud containing the concave area and the plane area in a direction parallel to the plane area;
in this embodiment, the three-dimensional point cloud acquisition module may be a device capable of acquiring three-dimensional point cloud data, such as a binocular camera.
The three-dimensional point cloud preprocessing module is connected with the plane area three-dimensional point cloud removing module and is used for preprocessing the three-dimensional point cloud to obtain a preprocessed three-dimensional point cloud, and the preprocessing operation comprises a downsampling operation and a filtering operation which are sequentially executed;
the planar area three-dimensional point cloud removing module is connected with the concave area three-dimensional point cloud dividing module and is used for taking a three-dimensional point with the largest depth value as a seed point in the three-dimensional point cloud of the non-planar area and obtaining an initial concave area three-dimensional point cloud by adopting an area growth method;
the method is also used for processing all three-dimensional points in the three-dimensional point cloud of the initial concave area by adopting an edge searching method to obtain a plurality of three-dimensional points to form the concave area edge;
and the three-dimensional point cloud of the concave area is obtained by collecting all three-dimensional points within the edge of the concave area.
Optionally, the three-dimensional point cloud preprocessing module, the planar area three-dimensional point cloud removing module and the concave area three-dimensional point cloud segmentation module are all intelligent mobile devices.
In this embodiment, in order to collect the three-dimensional point cloud of the concave area in real time, and process the three-dimensional point cloud of the concave area in real time to divide the same, the three-dimensional point cloud preprocessing module, the planar area three-dimensional point cloud removing module and the concave area three-dimensional point cloud dividing module are all arranged on the intelligent mobile device, and the intelligent mobile device can be a movable hardware device such as a computer or a mobile phone.
Optionally, the intelligent mobile device is a mobile phone.
In this embodiment, for more convenient practical use, the intelligent mobile device is a mobile phone, for example, when the pit area of the road is segmented, the road maintenance personnel can directly connect with the three-dimensional point cloud acquisition module (such as a binocular camera) by using the mobile phone, transmit the road three-dimensional point cloud acquired by the three-dimensional point cloud acquisition module to the mobile phone, and process by using the mobile phone APP to obtain the segmented pit area for subsequent processing.
Claims (7)
1. A three-dimensional point cloud segmentation method for segmenting a three-dimensional point cloud of a concave area from a three-dimensional point cloud containing the concave area and a plane area, the method comprising the following steps:
step 1, collecting three-dimensional point clouds containing a concave area and a plane area in a direction parallel to the plane area;
step 2, preprocessing the three-dimensional point cloud to obtain a preprocessed three-dimensional point cloud, wherein the preprocessing comprises a downsampling operation and a filtering operation which are sequentially executed;
step 3, deleting all three-dimensional points with depth values smaller than a depth threshold value according to the depth value of each three-dimensional point in the preprocessed three-dimensional point cloud to obtain a three-dimensional point cloud of a non-planar area, wherein the depth threshold value is twice the average value of the depth values of all three-dimensional points in the preprocessed three-dimensional point cloud;
step 4, taking the three-dimensional point with the largest depth value as a seed point in the three-dimensional point cloud of the non-planar area, and obtaining an initial three-dimensional point cloud of the concave area by adopting an area growth method;
step 5, processing all three-dimensional points in the three-dimensional point cloud of the initial concave area by adopting an edge searching method to obtain a plurality of three-dimensional points to form a concave area edge;
and 6, collecting all three-dimensional points within the edge of the concave area to obtain the concave area three-dimensional point cloud.
2. The three-dimensional point cloud segmentation method according to claim 1, wherein the step 2 is performed according to the following steps:
step 21, downsampling the three-dimensional point cloud by using a voxel grid filtering method to obtain the downsampled three-dimensional point cloud;
step 22, performing direct filtering on the down-sampled three-dimensional point cloud to obtain a direct filtered three-dimensional point cloud;
and step 23, carrying out statistical filtering on the three-dimensional point cloud after the direct filtering to obtain the processed three-dimensional point cloud.
3. The method of claim 2, wherein in the step 3, according to the depth value of each three-dimensional point in the preprocessed three-dimensional point cloud, deleting all three-dimensional points with depth values smaller than a depth threshold by using a RANSAC algorithm to obtain the three-dimensional point cloud of the non-planar area.
4. The method for three-dimensional point cloud segmentation according to claim 1, wherein when all three-dimensional points in the initial concave region three-dimensional point cloud are processed by adopting the edge search method, a vertex of the initial concave region three-dimensional point cloud is used as an initial boundary point, and a plurality of three-dimensional points are sequentially obtained to form a concave region edge by utilizing a boundary point maximum angle principle.
5. A three-dimensional point cloud segmentation device for implementing the three-dimensional point cloud segmentation method according to any one of claims 1 to 4, the device comprising a three-dimensional point cloud acquisition module, a three-dimensional point cloud preprocessing module, a planar area three-dimensional point cloud removal module and a recessed area three-dimensional point cloud segmentation module;
the three-dimensional point cloud acquisition module is connected with the three-dimensional point cloud preprocessing module and is used for acquiring the three-dimensional point cloud containing the concave area and the plane area in a direction parallel to the plane area;
the three-dimensional point cloud preprocessing module is connected with the plane area three-dimensional point cloud removing module and is used for preprocessing the three-dimensional point cloud to obtain a preprocessed three-dimensional point cloud, and the preprocessing operation comprises a downsampling operation and a filtering operation which are sequentially executed;
the planar area three-dimensional point cloud removing module is connected with the concave area three-dimensional point cloud dividing module and is used for taking a three-dimensional point with the largest depth value as a seed point in the three-dimensional point cloud of the non-planar area and obtaining an initial concave area three-dimensional point cloud by adopting an area growth method;
the method is also used for processing all three-dimensional points in the three-dimensional point cloud of the initial concave area by adopting an edge searching method to obtain a plurality of three-dimensional points to form the concave area edge;
and the three-dimensional point cloud of the concave area is obtained by collecting all three-dimensional points within the edge of the concave area.
6. The three-dimensional point cloud segmentation apparatus according to claim 5, wherein the three-dimensional point cloud preprocessing module, the planar area three-dimensional point cloud removing module and the concave area three-dimensional point cloud segmentation module are all intelligent mobile devices.
7. The three-dimensional point cloud segmentation apparatus according to claim 6, wherein the intelligent mobile device is a mobile phone.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811589435.9A CN109816664B (en) | 2018-12-25 | 2018-12-25 | Three-dimensional point cloud segmentation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811589435.9A CN109816664B (en) | 2018-12-25 | 2018-12-25 | Three-dimensional point cloud segmentation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109816664A CN109816664A (en) | 2019-05-28 |
| CN109816664B true CN109816664B (en) | 2023-05-09 |
Family
ID=66602476
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811589435.9A Active CN109816664B (en) | 2018-12-25 | 2018-12-25 | Three-dimensional point cloud segmentation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109816664B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110443836B (en) * | 2019-06-24 | 2022-03-18 | 中国人民解放军战略支援部队信息工程大学 | Point cloud data automatic registration method and device based on plane features |
| CN112166457A (en) * | 2019-08-29 | 2021-01-01 | 深圳市大疆创新科技有限公司 | Point cloud segmentation method and system and movable platform |
| CN110728688B (en) * | 2019-10-17 | 2022-02-22 | 杭州电子科技大学上虞科学与工程研究院有限公司 | Energy optimization-based three-dimensional mesh model segmentation method and system |
| CN110793437A (en) * | 2019-10-23 | 2020-02-14 | 珠海格力智能装备有限公司 | Positioning method and device of manual operator, storage medium and electronic equipment |
| CN112365582B (en) * | 2020-11-17 | 2022-08-16 | 电子科技大学 | Countermeasure point cloud generation method, storage medium and terminal |
| CN113723468B (en) * | 2021-08-06 | 2023-08-04 | 西南科技大学 | Object detection method of three-dimensional point cloud |
| CN115375699B (en) * | 2022-10-25 | 2023-03-24 | 杭州华橙软件技术有限公司 | Point cloud segmentation method, mobile robot and computer-readable storage medium |
| CN116716781A (en) * | 2023-06-06 | 2023-09-08 | 山东大学 | Intelligent pavement disease repairing device and method based on 3D printing |
| CN117115391B (en) * | 2023-10-24 | 2024-01-12 | 中科云谷科技有限公司 | Model updating method, device, computer equipment and computer readable storage medium |
| CN117994272B (en) * | 2024-04-07 | 2024-06-18 | 武汉科技大学 | Point cloud segmentation method and system applied to industrial disordered soft packet stack |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011070927A1 (en) * | 2009-12-11 | 2011-06-16 | 株式会社トプコン | Point group data processing device, point group data processing method, and point group data processing program |
| CN105989604A (en) * | 2016-02-18 | 2016-10-05 | 合肥工业大学 | Target object three-dimensional color point cloud generation method based on KINECT |
-
2018
- 2018-12-25 CN CN201811589435.9A patent/CN109816664B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN109816664A (en) | 2019-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109816664B (en) | Three-dimensional point cloud segmentation method and device | |
| CN109900338B (en) | Method and device for measuring volume of pavement pit | |
| CN109903327B (en) | Target size measurement method of sparse point cloud | |
| CN111724433B (en) | Crop phenotype parameter extraction method and system based on multi-view vision | |
| CN110717983B (en) | Building elevation three-dimensional reconstruction method based on knapsack type three-dimensional laser point cloud data | |
| CN111932688B (en) | Indoor plane element extraction method, system and equipment based on three-dimensional point cloud | |
| CN110340891B (en) | Mechanical arm positioning and grabbing system and method based on point cloud template matching technology | |
| CN106780524B (en) | Automatic extraction method for three-dimensional point cloud road boundary | |
| CN108133191A (en) | A kind of real-time object identification method suitable for indoor environment | |
| CN107481274B (en) | Robust reconstruction method of three-dimensional crop point cloud | |
| CN111696210A (en) | Point cloud reconstruction method and system based on three-dimensional point cloud data characteristic lightweight | |
| CN113628263A (en) | Point cloud registration method based on local curvature and neighbor characteristics thereof | |
| CN112785596B (en) | Dot cloud picture bolt segmentation and height measurement method based on DBSCAN clustering | |
| CN112200854B (en) | Leaf vegetable three-dimensional phenotype measuring method based on video image | |
| CN113362463A (en) | Workpiece three-dimensional reconstruction method based on Gaussian mixture model | |
| CN108280833B (en) | Skeleton extraction method for plant root system bifurcation characteristics | |
| Chang et al. | Automatic classification of lidar data into ground and non-ground points | |
| CN117706577A (en) | Ship size measurement method based on laser radar three-dimensional point cloud algorithm | |
| CN116051980B (en) | Building identification method, system, electronic equipment and medium based on oblique photography | |
| CN112365534B (en) | Large coal pile volume measurement method based on monocular camera three-dimensional reconstruction | |
| CN107808160B (en) | Three-dimensional building extraction method and device | |
| CN115456937A (en) | Deformation detection method directly based on TLS high-density point cloud measured value | |
| CN118314162B (en) | Dynamic visual SLAM method and device for time sequence sparse reconstruction | |
| CN115035184B (en) | Honey pomelo volume estimation method based on lateral multi-view reconstruction | |
| CN116188497B (en) | Method, device, equipment and storage medium for optimizing generation of DSM (digital image model) of stereo remote sensing image pair |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 710000 No. 605, west section, 6th floor, block B, operation and maintenance international headquarters building, No. 501, East Chang'an Street, national civil aerospace industry base, Xi'an, Shaanxi Province Applicant after: XI'AN ZHONGKE TIANTA TECHNOLOGY Co.,Ltd. Address before: Room 311, Zutong Building, No. 17 Information Avenue, New Industrial Park, High tech Zone, Xi'an City, Shaanxi Province, 710000 Applicant before: XI'AN ZHONGKE TIANTA TECHNOLOGY Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |