Full-waveform laser radar target recognition method and device with improved local G coefficient

Abstract

The present invention provides a full-waveform laser radar target recognition method and device with improved local G coefficient, the method comprising: obtaining and decomposing the echo waveform data received by the full-waveform laser radar to detect the target area to obtain point cloud data; taking each point of the point cloud data as the center point, taking the inverse of the Euclidean distance between the center point and its N neighboring points as the weight, calculating the improved local G coefficient of each point of the point cloud data; based on the improved local G coefficient, identifying the area where the suspected hidden target is located in the target area; clustering and segmenting the point cloud data in the area where the suspected hidden target is located, extracting the feature data of the suspected hidden target; inputting the feature data into a preset classifier to identify the suspected hidden target. The method can improve the existing local G coefficient and realize the efficient extraction of the suspected hidden target area.

Description

Technical Field

[0001] The present disclosure relates to the technical field of laser radar data processing, and in particular to a full-waveform laser radar target recognition method and device with improved local G coefficient. Background Art

[0002] The laser radar's emission pulse can penetrate the vegetation canopy to a certain extent, and can directly obtain target information under the vegetation cover. Compared with traditional optical cameras, it has incomparable advantages in the extraction of hidden targets. With the rapid development of full-waveform digital processing and storage technology, the full-waveform laser radar system can sample and record the waveform of the emission pulse and backscattered echo at a very small time interval. Among them, the echo waveform records the waveform information after the laser beam interacts with the objects at different height levels in the spot.

[0003] Compared with the discrete point cloud data directly obtained by the equipment, the full-waveform LiDAR data can not only increase the number of point clouds representing ground objects after waveform decomposition and feature extraction, but also obtain waveform features covering spatial and physical attribute information, which adds more data basis for subsequent ground object identification and classification, and also provides the possibility for the identification and extraction of hidden targets under vegetation cover. However, since hidden targets are hidden under vegetation cover, the echo of the laser radar pulse penetrating the vegetation is usually weak or undetectable, which is directly reflected in the point cloud data as the physical attribute information of the hidden target point cloud data is small or the spatial and physical attribute information is missing, which makes it more difficult to extract hidden targets. Therefore, more accurate detection of hidden target areas and more complete multi-feature analysis and classification algorithms are needed to achieve effective identification and extraction of hidden targets. Summary of the invention

[0004] In view of the above problems, the present invention provides a full-waveform lidar target recognition method and device with improved local G coefficient to improve the recognition accuracy of suspected hidden targets.

[0005] One aspect of the present disclosure provides a full-waveform laser radar target recognition method with an improved local G coefficient, comprising: acquiring and decomposing echo waveform data received by a full-waveform laser radar to detect a target area to obtain point cloud data; taking each point of the point cloud data as a center point and the inverse of the Euclidean distance between the center point and its N neighboring points as a weight, calculating an improved local G coefficient for each point of the point cloud data; identifying an area within the target area where a suspected hidden target is located based on the improved local G coefficient; clustering and segmenting the point cloud data within the area where the suspected hidden target is located to extract feature data of the suspected hidden target; and inputting the feature data into a preset classifier to identify the suspected hidden target.

[0006] Optionally, the formula for calculating the improved local G coefficient of each point of the point cloud data includes:

[0007]

[0008] Among them, G j (d) represents the local G coefficient of the jth center point, x i represents the number of components of the i-th point in the point cloud data, W ij (d) represents the reciprocal of the distance between the ith point and the jth center point.

[0009] Optionally, identifying the area in the target area where the suspected hidden target is located based on the improved local G coefficient includes: identifying the area where the points in the point cloud data where the improved local G coefficient is greater than a first preset threshold are located as the area where the suspected hidden target is located.

[0010] Optionally, it also includes: calculating the echo rate of the area where each point of the point cloud data is located respectively; identifying the area where the suspected hidden target is located within the target area based on the improved local G coefficient includes: identifying a point whose improved local G coefficient is greater than a first preset threshold and whose echo rate is less than a second preset threshold, and identifying the area where the point is located as the area where the suspected hidden target is located.

[0011] Optionally, respectively calculating the echo rate of the area where each point of the point cloud data is located includes: taking each point of the point cloud data as the center point, obtaining the number of first point clouds in a spherical area with the center point as the center of the sphere, and obtaining the number of second point clouds in a columnar area around an axis perpendicular to the ground where the center point is located, wherein the spherical area and the columnar area have the same radius, and the radius is twice the average point spacing; calculating the ratio of the first point cloud number to the second point cloud number to obtain the echo rate.

[0012] Optionally, clustering and segmenting the point cloud data within the area where the suspected hidden target is located to extract feature data of the suspected hidden target includes: clustering and segmenting the point cloud data within the area where the suspected hidden target is located to obtain point cloud data of the suspected hidden target; and extracting radiation features and geometric features of the point cloud data of the suspected hidden target.

[0013] Optionally, the classifier includes at least a support vector machine, a random forest, Adaboost, a naive Bayes, and a decision tree, and inputting the feature data into a preset classifier to identify the suspected hidden target includes: inputting the feature data into multiple classifiers, and selecting the best result from the recognition results of each classifier.

[0014] Optionally, after obtaining the point cloud data, the method includes: removing ground data of the target area from the point cloud data.

[0015] Optionally, it also includes: eliminating areas where suspected hidden targets are located that are beyond a preset elevation range.

[0016] Another aspect of the present disclosure provides a full-waveform laser radar target recognition device with improved local G coefficient, including: a data acquisition module, used to acquire and decompose the echo waveform data received by the full-waveform laser radar to detect the target area to obtain point cloud data; a coefficient calculation module, used to take each point of the point cloud data as the center point and the inverse of the Euclidean distance between the center point and its N neighboring points as the weight, and calculate the local G coefficient of each point of the point cloud data; an area identification module, used to identify the area where the suspected hidden target is located in the target area based on the local G coefficient; a feature extraction module, used to cluster and segment the point cloud data in the area where the suspected hidden target is located, and extract feature data of the suspected hidden target; a target identification module, used to input the feature data into a preset classifier to identify the suspected hidden target.

[0017] At least one of the above technical solutions adopted in the embodiments of the present disclosure can achieve the following beneficial effects:

[0018] The method provided by the present disclosure improves the existing local G coefficient by introducing the inverse distance weight, thereby achieving efficient extraction of suspected hidden target areas. Compared with the local G coefficient, the improved local G coefficient has higher accuracy in extracting suspected hidden target areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] For a more complete understanding of the present disclosure and its advantages, reference will now be made to the following description taken in conjunction with the accompanying drawings, in which:

[0020] Figure 1 The overall flow chart of a full-waveform laser radar target recognition method with improved local G coefficient provided by an embodiment of the present disclosure is schematically shown;

[0021] Figure 2 A schematic diagram showing the detailed process of a full-waveform laser radar target recognition method with improved local G coefficient provided by an embodiment of the present disclosure is shown;

[0022] Figure 3 The schematic diagram of echo rate calculation provided by the embodiment of the present disclosure is schematically shown;

[0023] Figure 4 The structural block diagram of a full-waveform laser radar target recognition device with improved local G coefficient provided by an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0024] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0025] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0026] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.

[0027] Some block diagrams and / or flow charts are shown in the accompanying drawings. It should be understood that some blocks or combinations thereof in the block diagrams and / or flow charts may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that these instructions, when executed by the processor, may create a device for implementing the functions / operations described in these block diagrams and / or flow charts.

[0028] Therefore, the technology of the present disclosure can be implemented in the form of hardware and / or software (including firmware, microcode, etc.). In addition, the technology of the present disclosure can take the form of a computer program product on a computer-readable medium storing instructions, which can be used by an instruction execution system or in combination with an instruction execution system. In the context of the present disclosure, a computer-readable medium can be any medium that can contain, store, transmit, propagate, or transmit instructions. For example, a computer-readable medium can include, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, device, or propagation medium. Specific examples of computer-readable media include: magnetic storage devices, such as magnetic tapes or hard disks (HDDs); optical storage devices, such as compact disks (CD-ROMs); memories, such as random access memories (RAMs) or flash memory; and / or wired / wireless communication links.

[0029] This paper proposes a full-waveform lidar target recognition method with improved local G coefficient. In view of the current problems in the recognition and extraction of hidden targets, based on an in-depth study of the full-waveform lidar waveform data acquisition mechanism and the characteristics of hidden target data, it combines target space statistical analysis, cluster segmentation, feature extraction, recognition extraction and other methods to efficiently realize the recognition and extraction of hidden targets under vegetation cover.

[0030] Figure 1 The overall flow chart of a full-waveform lidar target recognition method with improved local G coefficient provided by an embodiment of the present disclosure is schematically shown.

[0031] like Figure 1 As shown, the method provided by the embodiment of the present disclosure includes operations S110 to S150.

[0032] In operation S110, the echo waveform data received by the full-waveform laser radar when detecting the target area is acquired and decomposed to obtain point cloud data.

[0033] In operation S120, each point of the point cloud data is taken as a center point, and the inverse of the Euclidean distance between the center point and its N neighboring points is taken as a weight to calculate the local G coefficient of each point of the point cloud data.

[0034] In operation S130 , based on the local G coefficient, an area within the target area where a suspected hidden target is located is identified.

[0035] In operation S140, clustering and segmentation are performed on the point cloud data in the area where the suspected hidden target is located, and feature data of the suspected hidden target is extracted.

[0036] In operation S150, the feature data is input into a preset classifier to identify a suspected hidden target.

[0037] The method provided by the present invention is proposed based on the characteristics of high-value clustering of hidden target components, and the existing local G coefficient is improved to achieve efficient extraction of suspected hidden target areas. Compared with the local G coefficient, the improved local G coefficient has higher accuracy in extracting suspected hidden target areas, thereby improving the recognition rate of hidden targets in suspected hidden target areas.

[0038] The following is an explanation of the process details of the full-waveform lidar target recognition method with improved local G coefficient provided by the present invention.

[0039] Figure 2 A schematic diagram of the process details of a full-waveform lidar target recognition method with improved local G coefficient provided by an embodiment of the present disclosure is schematically shown.

[0040] like Figure 2As shown, in the embodiment of the present disclosure, the input data is point cloud data extracted after decomposition processing using the echo waveform data waveform.

[0041] After obtaining the point cloud data, the ground data of the target area is first removed from the point cloud data. Optionally, algorithms such as cloth simulation filtering (CSF) can be used to remove the ground data from the input data.

[0042] In view of the fact that hidden targets are often few in number and have incomplete spatial and physical property information in the detection area, it is necessary to first detect the areas in the detection area where hidden targets may exist, so as to improve the efficiency and accuracy of subsequent hidden target recognition and extraction. The present disclosure proposes an improved local G coefficient, which takes a certain laser point as the center and the inverse of the Euclidean distance between the point and its n nearest neighboring points as the weight (i.e., the inverse distance weight), calculates the improved local G coefficient of the point based on the component number feature, and then detects possible hidden target areas. The formula for calculating the improved local G coefficient of each point in the point cloud data includes:

[0043]

[0044] Among them, G j (d) represents the local G coefficient of the jth center point, x i Represents the number of components of the i-th point in the point cloud data, W ij (d) represents the inverse of the distance between the i-th point and the j-th center point.

[0045] In one implementation of the embodiment of the present disclosure, the area where the points in the point cloud data where the improved local G coefficient is greater than the first preset threshold are located can be identified as the area where the suspected hidden target is located.

[0046] In another implementation of the embodiment of the present disclosure, operation S160 may also be included.

[0047] In operation S160 , the echo rate of the area where each point of the point cloud data is located is calculated respectively.

[0048] Based on operation S160, the original operation S130, based on the improved local G coefficient, identifying the area where the suspected hidden target is located in the target area may also include:

[0049] Identify the point where the improved local G coefficient is greater than the first preset threshold and the echo rate is less than the second preset threshold, and identify the area where the point is located as the area where the suspected hidden target is located.

[0050] Based on the improved local G coefficient of spatial statistical analysis combined with the echo rate and the use of the elevation distribution of hidden targets, more accurate regional detection of suspected hidden targets can be achieved.

[0051] The echo rate refers to the ratio of the number of point clouds in a sphere with a radius of r and a certain laser point as the center to the number of point clouds in its vertical cylinder.

[0052] Figure 3 The following schematically shows a schematic diagram of echo rate calculation provided by an embodiment of the present disclosure.

[0053] like Figure 3 As shown, in this embodiment, each point of the point cloud data is taken as the center point, and the number of first point clouds in the spherical area with the center point as the sphere center is obtained, and the number of second point clouds in the columnar area around the axis perpendicular to the ground where the center point is located is obtained, wherein the radius of the spherical area and the columnar area is the same, and the radius is twice the average point spacing; the ratio of the number of the first point cloud to the number of the second point cloud is calculated to obtain the echo rate. The calculation formula of the echo rate is:

[0054]

[0055] Among them, ER represents the echo rate of a certain laser point, n 3D Indicates the number of point clouds in the sphere, n 2D Represents the number of point clouds in the cylinder, and n 3D ≤n 2D .

[0056] like Figure 3 As shown in the figure, if the radius r is too large, the transition zone between adjacent objects with different surface structures will be too large, which is not conducive to classification. Therefore, the most appropriate estimate of r is 2 times the average point spacing. Since the laser pulse cannot penetrate the building and the ground, there is only one echo reflected from the surface, while the vegetation is penetrable and will produce multiple echoes. Therefore, the building, the ground, and the exposed car have a high echo rate, while the echo rate of the hidden target under the cover of vegetation is low, so the hidden target exists in the area with low echo rate.

[0057] like Figure 2 As shown in the figure, after determining the area where the suspected hidden object is located based on the improved local G coefficient and echo rate, it is also necessary to eliminate the data outside the elevation range based on the approximate elevation range of the hidden target to further narrow the area where the suspected hidden object is located and improve the recognition accuracy.

[0058] After detecting the suspected hidden target area, it is necessary to extract the features of each suspected hidden target. Therefore, according to the characteristics of different detection areas and hidden targets, the clustering segmentation of suspected hidden targets is realized from a variety of point cloud clustering algorithms to obtain the point cloud data of suspected hidden targets and extract the radiation features and geometric features of the point cloud data of suspected hidden targets. Point cloud clustering algorithms can include classic algorithms such as Euclidean clustering, superbody clustering, density clustering, model clustering, hierarchical clustering, and spectral clustering.

[0059] Finally, the feature data is input into a plurality of pre-optimized classifiers, and the best result is selected from the recognition results of each classifier, and the classifiers include at least support vector machine, random forest, Adaboost, naive Bayes, decision tree, etc. In another embodiment of the present disclosure, the user can independently select the best classifier from the plurality of classifiers for target recognition, and complete the recognition and extraction of hidden targets. Compared with selecting one or several classifiers, the use of the machine learning classification library to autonomously select the best classifier reduces manual intervention, making the present disclosure more universal and having higher classification accuracy.

[0060] The full-waveform laser radar target recognition method based on the improved local G coefficient provided by the embodiment of the present disclosure improves the existing local G coefficient by introducing the inverse distance weight, thereby realizing the efficient extraction of suspected hidden target areas. Compared with the local G coefficient, the improved local G coefficient has a higher accuracy in extracting suspected hidden target areas, thereby improving the recognition rate of hidden targets.

[0061] Figure 4 The structural block diagram of a full-waveform laser radar target recognition device with improved local G coefficient provided by an embodiment of the present disclosure is schematically shown.

[0062] like Figure 4 As shown, another embodiment of the present disclosure provides a full-waveform lidar target recognition device 400 with improved local G coefficient, including: a data acquisition module 410, a coefficient calculation module 420, a region recognition module 430, a feature extraction module 440, and a target recognition module 450.

[0063] The data acquisition module 410 is used to acquire and decompose the echo waveform data received by the full-waveform laser radar to detect the target area, and obtain point cloud data;

[0064] The coefficient calculation module 420 is used to calculate the local G coefficient of each point in the point cloud data, taking each point in the point cloud data as the center point and the inverse of the Euclidean distance between the center point and its N neighboring points as the weight.

[0065] The region identification module 430 is used to identify the region where the suspected hidden target is located within the target region based on the local G coefficient.

[0066] The feature extraction module 440 is used to perform clustering and segmentation on the point cloud data in the area where the suspected hidden target is located, and extract feature data of the suspected hidden target.

[0067] The target identification module 450 is used to input the feature data into a preset classifier to identify the suspected hidden target. It is understandable that the data acquisition module 410, the coefficient calculation module 420, the region identification module 430, the feature extraction module 440, and the target identification module 450 can be combined in one module, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present invention, at least one of the data acquisition module 410, the coefficient calculation module 420, the region identification module 430, the feature extraction module 440, and the target identification module 450 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented in hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or in a suitable combination of software, hardware, and firmware. Alternatively, at least one of the data acquisition module 410, the coefficient calculation module 420, the region identification module 430, the feature extraction module 440, and the target identification module 450 can be at least partially implemented as a computer program module, and when the program is run by a computer, the function of the corresponding module can be executed.

[0068] The full waveform laser radar target recognition device 400 with improved local G coefficient provided by the embodiment of the present disclosure is similar to the Figures 1-2 The full-waveform lidar target recognition method with improved local G coefficient shown has the same technical features and can thus achieve the same technical effects, which will not be described in detail here.

[0069] It will be appreciated by those skilled in the art that the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways, even if such combinations and / or combinations are not explicitly described in the present disclosure. In particular, the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways without departing from the spirit and teachings of the present disclosure. All of these combinations and / or combinations fall within the scope of the present disclosure.

[0070] Although the present disclosure has been shown and described with reference to specific exemplary embodiments of the present disclosure, it should be understood by those skilled in the art that various changes in form and details may be made to the present disclosure without departing from the spirit and scope of the present disclosure as defined by the appended claims and their equivalents. Therefore, the scope of the present disclosure should not be limited to the above-mentioned embodiments, but should be determined not only by the appended claims, but also by the equivalents of the appended claims.

Claims

1. A full-waveform laser radar target recognition method with improved local G coefficient, characterized in that: include: Acquire and decompose the echo waveform data received by the full-waveform laser radar to detect the target area and obtain point cloud data; Taking each point of the point cloud data as the center point and the inverse of the Euclidean distance between the center point and its N neighboring points as the weight, the improved local G coefficient of each point of the point cloud data is calculated, wherein the formula for calculating the improved local G coefficient of each point of the point cloud data includes: ； in, Indicates The improved local G coefficient of the center point, Indicates the point cloud data The number of groups of points, Indicates Point and The reciprocal of the distance between the two center points; Based on the improved local G coefficient, identifying the area where the suspected hidden target is located in the target area; Clustering and segmenting the point cloud data in the area where the suspected hidden target is located to extract feature data of the suspected hidden target; The feature data is input into a preset classifier to identify the suspected hidden target.

2. The method according to claim 1, characterized in that The step of identifying the area where the suspected hidden target is located in the target area based on the improved local G coefficient includes: The area where the points in the point cloud data whose improved local G coefficient is greater than the first preset threshold are located is identified as the area where the suspected hidden target is located.

3. The method according to claim 1, characterized in that Also includes: Calculate the echo rate of each point of the point cloud data in the area where each point is located; The step of identifying the area where the suspected hidden target is located in the target area based on the improved local G coefficient includes: Identify the point where the improved local G coefficient is greater than the first preset threshold and the echo rate is less than the second preset threshold, and identify the area where the point is located as the area where the suspected hidden target is located.

4. The method according to claim 3, characterized in that The step of respectively calculating the echo rate of the area where each point of the point cloud data is located comprises: Taking each point of the point cloud data as the center point, respectively, obtaining the number of first point clouds in a spherical area with the center point as the sphere center, and obtaining the number of second point clouds in a columnar area around an axis perpendicular to the ground where the center point is located, wherein the spherical area and the columnar area have the same radius, which is twice the average point spacing; The ratio of the number of the first point cloud to the number of the second point cloud is calculated to obtain the echo rate.

5. The method according to claim 1, characterized in that The clustering and segmenting of the point cloud data in the area where the suspected hidden target is located to extract feature data of the suspected hidden target includes: Clustering and segmenting the point cloud data in the area where the suspected hidden target is located to obtain the point cloud data of the suspected hidden target; The radiation characteristics and geometric characteristics of the point cloud data of the suspected hidden target are extracted.

6. The method according to claim 1, characterized in that The classifier includes at least a support vector machine, a random forest, an Adaboost, a naive Bayes, and a decision tree. Inputting the feature data into a preset classifier to identify the suspected hidden target includes: The feature data is input into a plurality of the classifiers, and the best result is selected from the recognition results of the classifiers.

7. The method according to claim 1, characterized in that After obtaining the point cloud data, the following steps are included: The ground data of the target area is removed from the point cloud data.

8. The method according to claim 1 or 3, characterized in that: Also includes: Eliminate areas where suspected hidden targets are located that are beyond the preset elevation range.

9. A full-waveform laser radar target recognition device with improved local G coefficient, characterized in that: include: The data acquisition module is used to acquire and decompose the echo waveform data received by the full-waveform laser radar to detect the target area and obtain point cloud data; The coefficient calculation module is used to calculate the local G coefficient of each point of the point cloud data by taking each point of the point cloud data as the center point and the inverse of the Euclidean distance between the center point and its N neighboring points as the weight, wherein the formula for calculating the improved local G coefficient of each point of the point cloud data includes: ； in, Indicates The improved local G coefficient of the center point, Indicates the point cloud data The number of groups of points, Indicates Point and The reciprocal of the distance between the two center points; A region identification module, used for identifying a region where a suspected hidden target is located within the target region based on the local G coefficient; A feature extraction module is used to cluster and segment the point cloud data in the area where the suspected hidden target is located, and extract feature data of the suspected hidden target; The target recognition module is used to input the feature data into a preset classifier to identify the suspected hidden target.

Images