A multi-level denoising method and system for airborne laser scanning point cloud under severe weather conditions

Abstract

The present invention provides a multi-level denoising method and system for airborne laser scanning point clouds under severe weather conditions, and relates to the technical field of laser point cloud processing. The method comprises the following steps: Step 1, based on connected domain segmentation, quickly filter out clustered noise points far from the surface; Step 2, based on spatial density and connectivity analysis, adaptively filter out dense noise points diffusely around the surface; Step 3, based on guided filtering to smooth the front and rear offset distances, finely filter out residual discrete noise points around the surface. The method proposed in the present invention adopts a multi-level processing strategy, which can robustly remove complex noise points caused by severe weather conditions from coarse to fine. In addition, the method takes into account the spatial distribution prior of noise points, and can quickly identify clustered noise points far from the surface.

Description

Technical Field

[0001] The present invention relates to the technical field of laser point cloud processing, and in particular to a multi-level denoising method and system for airborne laser scanning point clouds under severe weather conditions. Background Art

[0002] Airborne laser scanning technology can quickly obtain high-precision three-dimensional point cloud data of the surface, and has been widely used in topographic mapping, forestry resource surveys, urban planning and other fields. Due to the influence of flying birds, multipath effects and bad weather, the original airborne laser point cloud usually contains a large number of noise points. In order to ensure the reliability of subsequent processing and application of point cloud data, it is usually necessary to filter out the noise points in the original point cloud in advance.

[0003] At present, there are a large number of methods for denoising three-dimensional point clouds, which can be mainly divided into three categories: (1) Denoising methods based on filtering principles. This type of method can effectively remove noise points, but may mistakenly smooth out the real detail information of the point cloud; (2) Denoising methods based on neighborhood spatial relationships. This type of method can effectively remove typical isolated noise points, but it is difficult to identify clustered dense noise points; (3) Denoising methods based on local density. This type of method can effectively remove sparse noise points, but usually cannot handle dense noise points.

[0004] In general, existing point cloud denoising methods can handle a small number of outlier high / low points in airborne laser scanning point clouds under good weather conditions. However, in bad weather conditions (such as wind, sand, fog, etc.), airborne laser scanning systems often produce a large number of dense and complex noise points, resulting in poor performance of existing denoising methods when processing such data.

[0005] Therefore, it is urgent to develop new robust filtering methods for complex noise in airborne laser scanning point clouds under severe weather conditions. Summary of the invention

[0006] Purpose of the invention: To propose a multi-level denoising method for airborne laser scanning point clouds under severe weather conditions, and further propose a system for implementing the above method, thereby effectively solving the above problems existing in the prior art.

[0007] First, a multi-level denoising method for airborne laser scanning point clouds under severe weather conditions is proposed, comprising the following steps:

[0008] Step 1: The acquired current laser scanning point cloud is used as the original point cloud, and the connected domain segmentation is performed on the original point cloud to filter out the clustered noise points far away from the surface, so as to obtain the first stage denoised point cloud data;

[0009] Step 2: Perform spatial density and connectivity analysis on the first-stage denoised point cloud data, filter out the dense noise points around the diffuse ground surface, and obtain the second-stage denoised point cloud data;

[0010] Step 3: For the second-stage denoised point cloud data, based on guided filtering to smooth the front and back offset distances, filter out the remaining discrete noise points around the ground surface, and output the final denoised point cloud data.

[0011] In a further embodiment of the first aspect, the process of obtaining the first-stage denoised point cloud data specifically includes:

[0012] Step 1.1: Based on connected component segmentation, divide the original point cloud into several sub-point clouds, and each sub-point cloud contains at least 10 points;

[0013] Step 1.2: Mark the sub-point cloud with the largest number of points as the sub-point cloud containing the ground surface, and directly regard other sub-point clouds as clustered noise points far from the ground surface and filter them out.

[0014] In a further embodiment of the first aspect, the process of obtaining the second-stage denoised point cloud data specifically includes:

[0015] Step 2.1: Traverse each point in the sub-point cloud containing the ground surface, and adaptively search for the neighborhood point set of each point;

[0016] Step 2.2: Count the number of points contained in the neighborhood point set, and regard the points with the number of neighborhood points less than a predetermined value as noise points and filter them out;

[0017] Step 2.3: Based on connected component segmentation, divide the remaining points into several sub-point clouds again, mark the sub-point cloud with the largest number of points among the several sub-point clouds as the sub-point cloud of the ground surface points that may contain residual noise points, and directly regard other sub-point clouds as dense noise points and filter them out. Each sub-point cloud contains at least 10 points.

[0018] In a further embodiment of the first aspect, the adaptive search radius R adaptive (P) of each point in Step 2.1 is calculated as follows:

[0019]

[0020] where α is the square of the distance product factor, represents the square of the minimum distance threshold, represents the square of the maximum distance threshold, and clamp(x,a,b) is a limiting function that returns x when x ∈ [a,b]; returns a when x < a; returns b when x > b.

[0021] In a further embodiment of the first aspect, the process of outputting the final denoised point cloud data specifically includes:

[0022] Step 3.1, loop through each point in the surface point cloud that may contain residual noise points, and search for a local neighborhood point set for each point;

[0023] Step 3.2, based on the spatial coordinates of all points in the local neighborhood point set of each point, calculate the smoothing matrix and offset vector, and update the coordinate position of the current point;

[0024] Step 3.3: traverse and calculate the offset distance of each point in the point cloud before and after smoothing, and treat the points with offset distance greater than the threshold as noise and remove them.

[0025] In a further embodiment of the first aspect, the expression of the local neighborhood point set of each point in step 3.1 is expressed in a matrix M:

[0026]

[0027] In the formula, (x j ,y j , z j ) represents the coordinates of neighborhood point j, and n represents the number of neighborhood points.

[0028] In a further embodiment of the first aspect, in step 3.2, the smoothing matrix A and the offset vector b are expressed as follows:

[0029] A=cov*(cov+I) -1

[0030] b=μ-A*μ

[0031] Where cov represents the covariance matrix of the local neighborhood point set matrix M; μ represents the mean vector of the local neighborhood point set matrix M;

[0032] Update the coordinate position of the current point as described in step 3.2. The calculation formula is as follows:

[0033] p ′ =A*p+b

[0034] Where p represents the coordinates of the current search point; p ′ Represents the coordinates of the current search point p after smooth update.

[0035] As a second aspect of the present invention, a multi-level denoising system for laser scanning point cloud is proposed, wherein the multi-level denoising system comprises a first-stage denoising unit, a second-stage denoising unit, and a third-stage denoising unit.

[0036] The first-stage denoising unit is used to use the acquired current laser scanning point cloud as the original point cloud, perform connected domain segmentation on the original point cloud, and filter out clustered noise points far away from the ground surface.

[0037] The second stage denoising unit performs spatial density and connectivity analysis to filter out dense noise points around the diffuse surface.

[0038] The third stage denoising unit smoothes the front and rear offset distances based on guided filtering to filter out the residual discrete noise points around the surface.

[0039] When the multi-level denoising system is operated, the multi-level denoising method for airborne laser scanning point clouds under severe weather conditions as disclosed in the first aspect and its further embodiments can be automatically executed.

[0040] As a third aspect of the present invention, a computer-readable storage medium is proposed, which stores at least one executable instruction. When the executable instruction is executed on an electronic device, the electronic device executes the multi-level denoising method for airborne laser scanning point clouds under severe weather conditions as disclosed in the first aspect and its further embodiments.

[0041] Compared with the existing methods, the present invention has the following beneficial effects:

[0042] 1. A multi-level processing strategy is adopted, which can robustly remove complex noise caused by severe weather conditions from coarse to fine.

[0043] 2. Taking into account the spatial distribution prior of noise points, clustered noise points far away from the surface can be quickly identified. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 Flow chart of a multi-level denoising method for airborne laser scanning point clouds under severe weather conditions in an embodiment of the present invention.

[0045] Figure 2 This is an example of cluster noise filtering based on connected domain segmentation by the method of the present invention.

[0046] Figure 3 This is an example of filtering out dense noise points around the surface based on spatial density and connectivity analysis using the method of the present invention.

[0047] Figure 4 This is an example of filtering out residual discrete noise points around the surface based on the offset distance before and after guided filtering smoothing by the method of the present invention.

[0048] Figure 5 Schematic diagram of the composition of a multi-level denoising system for laser scanning point cloud in an embodiment of the present invention. DETAILED DESCRIPTION

[0049] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the present invention. However, it will be apparent to one of ordinary skill in the art that the present invention may be practiced without one or more of these specific details. In other instances, well-known features of the art are not described in order to avoid obscuring the present invention.

[0050] Figure 1 Shown is a flowchart of a multi-level denoising method for airborne laser scanning point clouds under adverse weather conditions in this embodiment, including the following steps:

[0051] Step 1: Based on connected component segmentation, quickly filter out clustered noise points far from the ground surface. Figure 2 Shown is an example diagram of filtering out clustered noise points based on connected component segmentation, and each segmented block is marked with a rectangular frame. Step 1 specifically includes the following sub-steps:

[0052] Step 1.1: Use the connected component segmentation algorithm to segment the original point cloud into multiple sub-point clouds according to the spatial relationship between each point and its neighboring points, and each sub-point cloud contains at least 10 points;

[0053] Step 1.2: Considering that the data volume of the sub-point cloud containing the ground surface should be the largest, mark the sub-point cloud with the largest number of points as the "ground surface containing" sub-point cloud, and filter out other sub-point clouds as clustered noise points far from the ground surface.

[0054] Step 2: Based on spatial density and connectivity analysis, adaptively filter out discrete noise points densely distributed around the ground surface. Figure 3 Shown is an example diagram of filtering out dense noise points around the ground surface based on spatial density and connectivity analysis, and each segmented block is marked with a rectangular frame. Step 2 specifically includes the following sub-steps:

[0055] Step 2.1: Traverse each point in the "ground surface containing" point cloud, and search for the corresponding neighborhood point set according to the adaptive search radius. The adaptive search radius of each point is calculated as follows:

[0056]

[0057] In Formula 1, α is the square of the distance product factor, represents the square of the minimum distance threshold, represents the square of the maximum distance threshold. clamp(x,a,b) is a limiting function that returns x when x ∈ [a,b]; returns a when x < a; returns b when x > b;

[0058] Step 2.2: Determine whether the current point is noise according to the number of points contained in the neighborhood point set. If the number of points contained in the neighborhood point set of the current point is less than 3, it is regarded as noise and filtered out;

[0059] Step 2.3: Based on the connected domain segmentation, the remaining points are further divided into a series of sub-point sets (each sub-point set contains at least 10 points). The sub-point cloud with the largest number of points is marked as the surface point cloud that "may contain residual noise points", and the other sub-point clouds are regarded as dense noise points and filtered out.

[0060] Step 3: Smooth the front and rear offset distances based on guided filtering to finely filter out the residual discrete noise points around the surface. Figure 4 The following is an example of residual discrete noise removal based on the offset distance before and after the guided filter smoothing. Each segmentation block is marked with a rectangular frame. Step 3 specifically includes the following sub-steps:

[0061] Step 3.1, loop through each point in the surface point cloud that "may contain residual noise points" and search for the local neighborhood point set of each point with a radius of 1m;

[0062] Step 3.2, construct the neighborhood point set matrix M of each point:

[0063]

[0064] In formula 2, (x j ,y j , z j ) is the coordinate of the neighborhood point j, and n represents the number of neighborhood points.

[0065] According to the neighborhood point set matrix M, the mean vector μ and covariance matrix cov of the point set are calculated. The calculation formula is as follows:

[0066]

[0067] In formula 3, n represents the number of neighborhood points, (x j ,y j , z j ) represents the coordinates of the jth neighborhood point. In formula 4, p j represents the jth neighborhood point coordinate, μ is the average point coordinate, and n represents the number of neighborhood points.

[0068] According to the mean vector μ and the covariance matrix cov, the smoothing conversion matrix A and the offset vector b are further calculated:

[0069] A=cov*(cov+I) -1 (5)

[0070] b=μ-A*μ (6)

[0071] In formula 5, I is the unit matrix.

[0072] Update the current point coordinates. The calculation formula is as follows:

[0073] p， =A*p+b (7)

[0074] In formula 7, p is the coordinate of the current search point, p ， The coordinates after smooth update.

[0075] Step 3.3, calculate the offset distance of each point in the point cloud before and after smoothing:

[0076]

[0077] If d < 0.01, the current point is considered as noise filtered out.

[0078] As a preferred embodiment, a multi-level denoising system 400 for laser scanning point cloud is proposed. Figure 5 . The multi-level denoising system 400 includes a first-stage denoising unit 401, a second-stage denoising unit 402, and a third-stage denoising unit 403. The first-stage denoising unit 401 is used to use the current laser scanning point cloud acquired as the original point cloud, perform connected domain segmentation on the original point cloud, filter out clustered noise points far away from the surface, and send it to the second-stage denoising unit 402 after processing. The second-stage denoising unit 402 performs spatial density and connectivity analysis to filter out dense noise points diffusely around the surface, and sends it to the third-stage denoising unit 403 after processing. The third-stage denoising unit 403 smoothes the front and rear offset distances based on the guided filter to filter out the residual discrete noise points around the surface. The processing process of the above-mentioned first-stage denoising unit 401, the second-stage denoising unit 402, and the third-stage denoising unit 403 is as disclosed in the above-mentioned embodiment, and will not be repeated here.

[0079] The multi-level denoising method for airborne laser scanning point cloud under severe weather conditions disclosed in the above embodiment can be written into a set of executable instructions for its operation logic in the actual application process, and the executable instructions are written into a storage medium and run on an electronic device. More specific examples of the computer-readable storage medium mentioned in this embodiment may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. The computer-readable storage medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The readable signal medium may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, device, or device.

[0080] As described above, although the present invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the present invention itself. Various changes may be made to it in form and detail without departing from the spirit and scope of the present invention as defined in the appended claims.

Claims

1. A multi-level denoising method for airborne laser scanning point clouds under severe weather conditions, characterized in that: The steps include: Step 1: The acquired current laser scanning point cloud is used as the original point cloud, and the connected domain segmentation is performed on the original point cloud to filter out the clustered noise points far away from the surface, so as to obtain the first stage denoised point cloud data; Step 1.1, based on connected domain segmentation, divide the original point cloud into several sub-point clouds; Step 1.2, mark the sub-point cloud with the largest number of points as the sub-point cloud containing the ground surface, and directly regard the other sub-point clouds as clustered noise points far away from the ground surface and filter them out; Step 2: Perform spatial density and connectivity analysis on the first-stage denoised point cloud data to filter out dense noise points around the diffuse surface to obtain the second-stage denoised point cloud data; Step 2.1, traverse each point in the surface sub-point cloud, and adaptively search for the neighborhood point set of each point; Step 2.2, counting the number of points contained in the neighborhood point set, and treating points with a number of neighborhood points less than a predetermined value as noise points and filtering them out; Step 2.3, based on the connected domain segmentation, the remaining points are divided into several sub-point clouds again, and the sub-point cloud with the largest number of points among the sub-point clouds is marked as a surface point cloud that may contain residual noise points, and the other sub-point clouds are directly regarded as dense noise points and filtered out; Step 3: For the denoised point cloud data in the second stage, the offset distances before and after are smoothed based on the guided filter, the residual discrete noise points around the surface are filtered out, and the final denoised point cloud data is output; Step 3.1, loop through each point in the surface point cloud that may contain residual noise points, and search for the local neighborhood point set of each point; the expression of the local neighborhood point set of each point is expressed by matrix M: ； In the formula, represents the coordinates of neighborhood point j, and n represents the number of neighborhood points; Step 3.2, based on the spatial coordinates of all points in the local neighborhood point set of each point, calculate the smoothing matrix and offset vector, and update the coordinate position of the current point; The expressions of the smoothing matrix A and the offset vector b are as follows: ； ； In the formula, Represents the covariance matrix of the local neighborhood point set matrix M; Represents the mean vector of the local neighborhood point set matrix M; I is the identity matrix; The calculation formula for updating the coordinate position of the current point is as follows: ； In the formula, p represents the coordinates of the current search point; Represents the coordinates of the current search point p after smooth update; Step 3.3: traverse and calculate the offset distance of each point in the point cloud before and after smoothing, and treat the points with offset distance greater than the threshold as noise and remove them.

2. The multi-level denoising method for airborne laser scanning point cloud under severe weather conditions according to claim 1 is characterized in that: Each sub-point cloud contains at least 10 points.

3. The multi-level denoising method for airborne laser scanning point cloud under severe weather conditions according to claim 1 is characterized in that: Adaptive search radius for each point in step 2.1 The calculation formula is as follows: ； In the formula, is the square of the distance multiplication factor, represents the square of the minimum distance threshold, represents the square of the maximum distance threshold, is a limiting function, when Return when Return ;when Returns b.

4. A multi-level denoising system for laser scanning point clouds, characterized in that: include: The first stage denoising unit is used to use the acquired current laser scanning point cloud as the original point cloud, perform connected domain segmentation on the original point cloud, and filter out clustered noise points far away from the ground surface; The second stage denoising unit performs spatial density and connectivity analysis to filter out dense noise points around the diffuse surface; The third stage denoising unit smoothes the front and rear offset distances based on guided filtering to filter out the residual discrete noise points around the surface; When the multi-level denoising system is operated, the multi-level denoising method for airborne laser scanning point cloud under severe weather conditions as described in any one of claims 1 to 3 can be automatically executed.

5. A computer-readable storage medium, characterized in that: The storage medium stores at least one executable instruction, and when the executable instruction is executed on the electronic device, the electronic device executes the multi-level denoising method for airborne laser scanning point clouds under severe weather conditions as described in any one of claims 1 to 3.

Images