Automatic identification method and system for a laser scanning point cloud target ball

Abstract

The application provides an automatic identification method and system for a laser scanning point cloud target ball, relates to the technical field of target automatic identification, and solves the problems of low target identification accuracy and slow target search speed in the existing automatic identification method for a laser scanning point cloud spherical target, acquires three-dimensional laser scanning point cloud data of the target, converts the three-dimensional laser scanning point cloud data into a depth map, performs edge enhancement filtering on the depth map to obtain a filtered image, identifies and extracts a circle in the filtered image, marks the circle as a target ball circle, filters the target ball circle to obtain a qualified target ball circle, pre-processes the qualified target ball circle to obtain target ball point cloud, performs target ball model fitting on the target ball point cloud, and retains a qualified target ball model according to a fitting result, so as to take the target ball model as an identification result of the target ball. The application effectively improves the target identification accuracy and the target search speed.

Description

Technical Field

[0001] This invention relates to the technical field of automatic target recognition, and in particular to an automatic recognition method and system for laser-scanned point cloud target spheres. Background Technology

[0002] Laser scanning technology is a technique that uses a laser to emit a laser beam and a photoelectric sensor to receive the reflected laser to measure information such as the shape, size, and position of an object's surface. Laser scanning technology can accurately capture the three-dimensional coordinate data of an object's surface to generate high-precision point cloud models or 3D reconstructions. Therefore, the accuracy and reliability of point cloud data are crucial in the laser scanning process.

[0003] Currently, there are three main methods for identifying target spheres in point clouds: direct target sphere search based on point clouds, target sphere extraction based on reflection intensity, and target sphere extraction based on RGB information. Direct target sphere search based on point clouds involves analyzing the geometric shape and features in the point cloud data to determine the presence of a target sphere and then automatically extracting it. However, this method has low extraction efficiency, is unstable, and is easily affected by scene interference. Target sphere extraction based on reflection intensity utilizes reflection intensity information obtained from a laser scanner, analyzing the intensity distribution characteristics in the point cloud data to determine the presence of a target sphere and then automatically extracting it. However, this method has limitations, such as data lacking reflection intensity information or in some scenes where the reflection intensity contrast is not obvious, leading to inaccurate identification results. Target sphere extraction based on RGB information utilizes RGB color information from point cloud data collected by a laser scanner, analyzing color features to determine the presence of a target sphere and then automatically extracting it. However, this method sometimes encounters situations where RGB information is missing or the shooting conditions are not met, thus affecting the accuracy of target sphere extraction and identification.

[0004] Existing technology discloses an automatic identification method for spherical targets from ground-based 3D laser scanning point clouds. First, during the conversion of the raw data scanned by the laser scanner from 3D polar coordinates to 3D rectangular coordinates, the distance and angle information in the polar coordinates are retained. Then, angle, distance, and grayscale information are used to achieve rapid coarse identification of spherical targets. Next, the spherical target data is fitted based on its 3D rectangular coordinates. The tolerance of the fitted sphere is combined with information such as its standard deviation, radius, and center to determine whether the target is truly a sphere. However, relying on angle, distance, and grayscale information to identify spherical targets in the coarse identification process has two drawbacks. First, this search method is not accurate enough for complex target shapes or background interference, leading to reduced target identification accuracy and incorrect results. Second, the complexity of angle, distance, and grayscale information results in slow target search speed, i.e., slow judgment speed. Summary of the Invention

[0005] To address the issues of low target recognition accuracy and slow target search speed in existing automatic target recognition methods for laser-scanned point clouds, this invention proposes an automatic target sphere recognition method and system for laser-scanned point clouds, which effectively improves target recognition accuracy and target search speed.

[0006] To achieve the above-mentioned technical effects, the technical solution of the present invention is as follows:

[0007] An automatic identification method for a target sphere in a laser-scanned point cloud includes the following steps:

[0008] S1. Acquire the 3D laser scanning point cloud data of the target;

[0009] S2. Convert the 3D laser scanning point cloud data into a depth map;

[0010] S3. Apply edge enhancement filtering to the depth map to obtain the filtered image;

[0011] S4. Identify and extract circles in the filtered image, and denote the circles as target sphere circles;

[0012] S5. Filter the target ball circle to obtain a qualified target ball circle;

[0013] S6. Preprocess the qualified target sphere circles to obtain the target sphere point cloud;

[0014] S7. Fit the target sphere model to the target sphere point cloud and determine whether the fitting result of the target sphere model is greater than the residual threshold. If so, remove the unqualified target sphere model; otherwise, retain the qualified target sphere model and use it as the target sphere recognition result.

[0015] Preferably, the process of converting the three-dimensional laser scanning point cloud data into a depth map includes:

[0016] S21. Select the location of the station center as the reference point, and calculate the distance from each point in the 3D laser scanning point cloud data to the station center;

[0017] S22. Calculate the horizontal and vertical angles from each point in the 3D laser scanning point cloud data to the center of the station;

[0018] S23. Combine the distance, horizontal angle, and pitch angle of each point in the 3D laser scanning point cloud data to the center of the station to obtain the depth information of each point in the 3D laser scanning point cloud data, and generate a depth map P.

[0019] Preferably, the process of performing edge enhancement filtering on the depth map includes:

[0020] S31. Perform a Laplacian transform on the depth map to generate an image with enhanced edges;

[0021] S32. Apply Gaussian filtering to the edge-enhanced image, and then use the Canny operator to perform edge detection on the Gaussian-filtered image to generate a filtered image.

[0022] Preferably, the process of identifying and extracting circles in the filtered image includes: performing circle identification and extraction in the filtered image through Hough transform to obtain a first set of circles Y, denoted as target sphere circles, and the radius of the circles is the radius Rt0 of the target sphere circles.

[0023] Preferably, the target ball circle is filtered, and the specific process includes:

[0024] S51. Using the coordinates of the center pixel within the target sphere circle, obtain the depth value of that center pixel in the depth map P, and convert the obtained depth value into the distance L0 from the target sphere surface point to the station center.

[0025] S52. Based on the distance L0, calculate the distance L from the center of the target sphere to the center of the station. The specific calculation formula is as follows:

[0026] L = L0 + R

[0027] Where R represents the radius of the target sphere;

[0028] S53. Calculate the three-dimensional coordinates Pmc of the target sphere's center based on the distance L from the target sphere's center to the station's center and the coordinates of the center pixel within the target sphere's circle.

[0029] S54. Calculate the theoretical pixel radius Rt1 of the target sphere whose center is located at the three-dimensional coordinate Pmc in the depth map P;

[0030] S55. Determine whether the difference between Rt0 and Rt1 is greater than the radius threshold. If so, record the target ball circle with the current radius Rt0 as an unqualified target ball circle and filter out the unqualified target ball circles in the first circular set Y. Otherwise, record the target ball circle with the current radius Rt0 as a qualified target ball circle, retain the qualified target ball circles, and generate the second circular set Y0.

[0031] Preferably, the qualified target ball circles are pre-processed, and the specific process includes:

[0032] S61. In the second circular set Y0, the coordinates of the center pixel point inside each target ball circle in the second circular set Y0 are converted into the three-dimensional coordinates Pmc of the center of the target ball;

[0033] S62. The three-dimensional coordinates Pmc of the center of the target sphere described in S61 are denoted as the target sphere point cloud, and the target sphere point cloud is used to form a target sphere point cloud set PC.

[0034] Preferably, the process of fitting a target sphere model to the target sphere point cloud and determining whether the fitting result of the target sphere model is greater than the residual threshold includes:

[0035] S71. Construct a target sphere model based on the three-dimensional coordinates Pmc of the center of each target sphere in the target sphere point cloud set PC and the radius R of each target sphere;

[0036] S72. Calculate the distance residual rms from each point on the target ball point cloud in the target ball point cloud set PC to the corresponding target ball model surface, and denote the distance residual rms as the fitting result;

[0037] S73. Determine whether the distance residual rms is greater than the residual threshold. If so, the fitting result of the target ball model is greater than the residual threshold, and remove the unqualified target ball model from the second circular set Y0; otherwise, leave the qualified target ball model in the second circular set Y0 and use it as the identification result of the target ball.

[0038] Preferably, the three-dimensional laser scanning point cloud data is obtained by scanning the target using a three-dimensional laser scanner.

[0039] This invention also proposes an automatic identification system for laser-scanned point cloud target spheres, the system comprising:

[0040] The acquisition module is used to acquire the three-dimensional laser scanning point cloud data of the target;

[0041] The conversion module is used to convert 3D laser scanning point cloud data into depth maps;

[0042] The enhancement filtering module is used to perform edge enhancement filtering on the depth map to obtain a filtered image.

[0043] The identification and extraction module is used to identify and extract circles in the filtered image, and the circles are referred to as target spheres.

[0044] A filtering module is used to filter the target ball circles to obtain qualified target ball circles;

[0045] The preprocessing module is used to preprocess qualified target sphere circles to obtain target sphere point clouds;

[0046] The judgment module is used to fit the target sphere model to the target sphere point cloud and determine whether the fitting result of the target sphere model is greater than the residual threshold. If so, the unqualified target sphere model is removed; otherwise, the qualified target sphere model is retained and used as the recognition result of the target sphere.

[0047] Preferably, a computer device is characterized by comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus;

[0048] The memory is used to store at least one executable instruction that causes the processor to perform the operation of the automatic identification method for laser scanning point cloud target spheres as described above.

[0049] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0050] This invention proposes an automatic target sphere identification method and system based on laser scanning point cloud data. First, it acquires 3D laser scanning point cloud data of the target, converts this data into a depth map, and applies edge enhancement filtering to the depth map to obtain a filtered image. The purpose of this method is to coarsely extract the target sphere using the depth map derived from the point cloud data, reducing interference from external environments such as scene noise. Then, it identifies and extracts circles from the filtered image, filtering these circles. The filtered, qualified target sphere circles undergo preprocessing. Finally, it fits the preprocessed target sphere point cloud data, selecting qualified target sphere models based on the fitting results to complete target sphere identification. The fitted target sphere models exhibit high robustness. This invention, through a coarse extraction method using a depth map converted from point cloud data, combined with edge enhancement filtering, identification and extraction, filtering, preprocessing, and model fitting, removes unqualified target sphere models, improving target identification accuracy and search speed, and enhancing the robustness of the target sphere models. Attached Figure Description

[0051] Figure 1 This is a flowchart illustrating an automatic identification method for laser scanning point cloud target spheres proposed in an embodiment of the present invention.

[0052] Figure 2 This represents an image of Laplacian edge enhancement proposed in an embodiment of the present invention;

[0053] Figure 3 This represents the filtered image extracted by the Canny operator proposed in this embodiment of the invention;

[0054] Figure 4 This diagram illustrates the relevant parameters of the A neural network proposed in this embodiment of the invention.

[0055] Figure 5 This represents the qualified target ball model left behind in the embodiments of the present invention.

[0056] Figure 6 This is a structural diagram illustrating an automatic identification system for laser-scanned point cloud target spheres proposed in an embodiment of the present invention.

[0057] Figure 7 This is a schematic diagram of the structure of a computer device proposed in an embodiment of the present invention. Detailed Implementation

[0058] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent.

[0059] To better illustrate this embodiment, some parts of the accompanying drawings may be omitted, enlarged, or reduced, and do not represent actual dimensions. The descriptions of directions such as "up" and "down" are not intended to limit this patent.

[0060] It is understandable to those skilled in the art that some well-known details may be omitted from the accompanying drawings;

[0061] The positional relationships depicted in the accompanying drawings are for illustrative purposes only and should not be construed as limiting this patent.

[0062] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0063] Example 1

[0064] like Figure 1 As shown in the figure, this embodiment proposes an automatic identification method for laser scanning point cloud target spheres, including the following steps:

[0065] S1. Acquire the 3D laser scanning point cloud data of the target;

[0066] In step S1, the three-dimensional laser scanning point cloud data is obtained by scanning the target with a three-dimensional laser scanner;

[0067] S2. Convert the 3D laser scanning point cloud data into a depth map;

[0068] In step S2, the process of converting the three-dimensional laser scanning point cloud data into a depth map specifically includes:

[0069] S21. Select the location of the station center as the reference point, and calculate the distance from each point in the 3D laser scanning point cloud data to the station center; where the coordinates of the station center are defined as (0,0,0), and the radius of the target sphere is R;

[0070] S22. Calculate the horizontal and vertical angles from each point in the 3D laser scanning point cloud data to the center of the station;

[0071] S23. Combine the distance, horizontal angle, and pitch angle of each point in the 3D laser scanning point cloud data to the center of the station to obtain the depth information of each point in the 3D laser scanning point cloud data, and generate a depth map P.

[0072] S3. Apply edge enhancement filtering to the depth map to obtain the filtered image;

[0073] In step S3, the edge enhancement filtering of the depth map specifically includes:

[0074] S31. Perform a Laplacian transform on the depth map. (See below) Figure 2 Generate an edge-enhanced image P0;

[0075] S32. Apply Gaussian filtering to the edge-enhanced image, and then use the Canny operator to perform edge detection on the Gaussian-filtered image. See [link to relevant documentation]. Figure 3 Generate a filtered image P1.

[0076] S4. Identify and extract circles in the filtered image, and denote the circles as target sphere circles;

[0077] In step S4, the process of identifying and extracting circles from the filtered image specifically includes: performing circle identification and extraction in the filtered image using Hough transform, see [link to relevant documentation]. Figure 4 We obtain the first set of circles Y, and denote the circles in the first set of circles Y as the target sphere circles, and the radius of the circles as the radius Rt0 of the target sphere circles.

[0078] S5. Filter the target ball circle to obtain a qualified target ball circle;

[0079] In step S5, the first set of circles Y is traversed, and the target sphere circles are filtered. The specific process includes:

[0080] S51. Using the coordinates (u,v) of the center pixel point inside the target sphere circle, obtain the depth value of the center pixel point in the depth map P, and convert the obtained depth value into the distance L0 from the target sphere surface point to the center of the station.

[0081] S52. Based on the distance L0, calculate the distance L from the center of the target sphere to the center of the station. The specific calculation formula is as follows:

[0082] L = L0 + R

[0083] Where R represents the radius of the target sphere;

[0084] S53. Calculate the three-dimensional coordinates Pmc of the target sphere's center based on the distance L from the target sphere's center to the station's center and the coordinates (u,v) of the center pixel within the target sphere's circle.

[0085] S54. Calculate the theoretical pixel radius Rt1 of the target sphere whose center is located at the three-dimensional coordinate Pmc in the depth map P;

[0086] S55. Determine whether the difference between Rt0 and Rt1 is greater than the radius threshold. If so, record the target ball circle with the current radius Rt0 as an unqualified target ball circle and filter out the unqualified target ball circles in the first circular set Y. Otherwise, record the target ball circle with the current radius Rt0 as a qualified target ball circle, retain the qualified target ball circles, and generate the second circular set Y0.

[0087] S6. Preprocess the qualified target sphere circles to obtain the target sphere point cloud;

[0088] In step S6, the qualified target ball circles are preprocessed, and the specific process includes:

[0089] S61. In the second circular set Y0, the coordinates of the center pixel point inside each target ball circle in the second circular set Y0 are converted into the three-dimensional coordinates Pmc of the center of the target ball;

[0090] S62. The three-dimensional coordinates Pmc of the center of the target sphere described in S61 are denoted as the target sphere point cloud, and the target sphere point cloud is used to form a target sphere point cloud set PC.

[0091] S7. Fit the target sphere model to the target sphere point cloud and determine whether the fitting result of the target sphere model is greater than the residual threshold. If so, remove the unqualified target sphere model; otherwise, retain the qualified target sphere model and use it as the target sphere recognition result.

[0092] In step S7, the process of fitting the target sphere point cloud with a target sphere model and determining whether the fitting result of the target sphere model is greater than the residual threshold includes:

[0093] S71. Construct a target sphere model based on the three-dimensional coordinates Pmc of the center of each target sphere in the target sphere point cloud set PC and the radius R of each target sphere;

[0094] In step S71, the constructed target sphere models are used to form a target sphere model set M;

[0095] S72. Calculate the distance residual rms from each point on the target ball point cloud in the target ball point cloud set PC to the corresponding target ball model surface, and denote the distance residual rms as the fitting result;

[0096] In step S72, each calculated distance residual rms is used to form the corresponding standard deviation set RMS; that is, the distance residual rms belongs to a subset of the standard deviation set RMS.

[0097] S73. Determine if the distance residual RMS is greater than the residual threshold. If so, the fitting result of the target ball model is greater than the residual threshold, and remove the unqualified target ball model from the second circular set Y0; otherwise, retain the qualified target ball model in the second circular set Y0 and use it as the target ball identification result, and the automatic target ball search is completed. See also Figure 5 Ultimately, the three correct target balls were identified.

[0098] In this embodiment, firstly, 3D laser scanning point cloud data of the target is acquired. This data is then converted into a depth map, and edge enhancement filtering is applied to the depth map to obtain a filtered image. The purpose of this is to coarsely extract the target sphere using the depth map derived from the point cloud data, reducing interference from external environments such as scene disturbances. Next, circles are identified and extracted from the filtered image, and these circles are filtered. The qualified target sphere circles are then preprocessed. Finally, the preprocessed target sphere point cloud is fitted, and qualified target sphere models are selected based on the fitting results to complete target sphere recognition. The fitted target sphere model exhibits high robustness. This invention, through a coarse extraction method using a depth map converted from point cloud data, combined with edge enhancement filtering, identification and extraction, filtering, preprocessing, and model fitting, removes unqualified target sphere models, improving target recognition accuracy and target search speed, and enhancing the robustness of the target sphere model.

[0099] Example 2

[0100] See Figure 6 This embodiment proposes an automatic identification system 600 for laser scanning point cloud target spheres, the system comprising:

[0101] The acquisition module 601 is used to acquire the three-dimensional laser scanning point cloud data of the target;

[0102] The conversion module 602 is used to convert 3D laser scanning point cloud data into depth maps;

[0103] The enhancement filtering module 603 is used to perform edge enhancement filtering on the depth map to obtain a filtered image;

[0104] The identification and extraction module 604 is used to identify and extract circles in the filtered image, and the circles are referred to as target sphere circles.

[0105] The filter module 605 is used to filter the target ball circle to obtain a qualified target ball circle.

[0106] The preprocessing module 606 is used to preprocess the qualified target sphere circles to obtain the target sphere point cloud;

[0107] The judgment module 607 is used to fit the target sphere model to the target sphere point cloud and determine whether the fitting result of the target sphere model is greater than the residual threshold. If so, the unqualified target sphere model is removed; otherwise, the qualified target sphere model is retained and used as the recognition result of the target sphere.

[0108] In this embodiment, firstly, 3D laser scanning point cloud data of the target is acquired. This data is then converted into a depth map, and edge enhancement filtering is applied to the depth map to obtain a filtered image. The purpose of this is to coarsely extract the target sphere using the depth map derived from the point cloud data, reducing interference from external environments such as scene disturbances. Next, circles are identified and extracted from the filtered image, and these circles are filtered. The qualified target sphere circles are then preprocessed. Finally, the preprocessed target sphere point cloud is fitted, and qualified target sphere models are selected based on the fitting results to complete target sphere recognition. The fitted target sphere model exhibits high robustness. This invention, through a coarse extraction method using a depth map converted from point cloud data, combined with edge enhancement filtering, identification and extraction, filtering, preprocessing, and model fitting, removes unqualified target sphere models, improving target recognition accuracy and target search speed, and enhancing the robustness of the target sphere model.

[0109] Example 3

[0110] This embodiment presents a computer device, see [link to documentation]. Figure 7 It includes: a processor 701, a memory 702, a communication interface 703, and a communication bus 704, wherein the processor 701, the memory 702, and the communication interface 703 communicate with each other through the communication bus 704;

[0111] The processor 701, memory 702, and communication interface 703 communicate with each other via communication bus 704. Communication interface 703 is used to communicate with other network elements, such as clients or other servers. The processor 701 executes executable instructions 705, specifically performing the relevant steps in the above-described embodiment of the automatic identification method for laser-scanned point cloud target spheres.

[0112] Specifically, the executable instructions 705 may include program code. The processor 701 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The computer device may include one or more processors of the same type, such as one or more CPUs; or it may include processors of different types, such as one or more CPUs and one or more ASICs.

[0113] Memory 702 is used to store executable instructions 705. Memory 702 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0114] Executable instruction 705 can be invoked by processor 701 to cause the computer device to perform the following operations:

[0115] S1. Acquire the 3D laser scanning point cloud data of the target;

[0116] S2. Convert the 3D laser scanning point cloud data into a depth map;

[0117] S3. Apply edge enhancement filtering to the depth map to obtain the filtered image;

[0118] S4. Identify and extract circles in the filtered image, and denote the circles as target sphere circles;

[0119] S5. Filter the target ball circle to obtain a qualified target ball circle;

[0120] S6. Preprocess the qualified target sphere circles to obtain the target sphere point cloud;

[0121] S7. Fit the target sphere model to the target sphere point cloud and determine whether the fitting result of the target sphere model is greater than the residual threshold. If so, remove the unqualified target sphere model; otherwise, retain the qualified target sphere model and use it as the target sphere recognition result.

[0122] In this embodiment, firstly, 3D laser scanning point cloud data of the target is acquired. This data is then converted into a depth map, and edge enhancement filtering is applied to the depth map to obtain a filtered image. The purpose of this is to coarsely extract the target sphere using the depth map derived from the point cloud data, reducing interference from external environments such as scene disturbances. Next, circles are identified and extracted from the filtered image, and these circles are filtered. The qualified target sphere circles are then preprocessed. Finally, the preprocessed target sphere point cloud is fitted, and qualified target sphere models are selected based on the fitting results to complete target sphere recognition. The fitted target sphere model exhibits high robustness. This invention, through a coarse extraction method using a depth map converted from point cloud data, combined with edge enhancement filtering, identification and extraction, filtering, preprocessing, and model fitting, removes unqualified target sphere models, improving target recognition accuracy and target search speed, and enhancing the robustness of the target sphere model.

[0123] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. An automatic identification method for target spheres in laser scanning point clouds, characterized in that, Includes the following steps: S1. Acquire the 3D laser scanning point cloud data of the target; S2. Convert the 3D laser scanning point cloud data into a depth map; S3. Apply edge enhancement filtering to the depth map to obtain the filtered image; S4. Identify and extract circles in the filtered image, and denote the circles as target sphere circles; S5. Filter the target ball circles to obtain qualified target ball circles. The specific process includes: S51. Using the coordinates of the center pixel within the target sphere circle, obtain the depth value of that center pixel in the depth map P, and convert the obtained depth value into the distance L0 from the target sphere surface point to the station center. S52. Based on the distance L0, calculate the distance L from the center of the target sphere to the center of the station. The specific calculation formula is as follows: L=LO+R Where R represents the radius of the target sphere; S53. Calculate the three-dimensional coordinates Pmc of the target sphere's center based on the distance L from the target sphere's center to the station's center and the coordinates of the center pixel within the target sphere's circle. S54. Calculate the theoretical pixel radius Rt1 of the target sphere whose center is located at the three-dimensional coordinate Pmc in the depth map P; S55. Determine whether the difference between Rt0 and Rt1 is greater than the radius threshold. If so, record the target ball circle with the current radius Rt0 as an unqualified target ball circle and filter out the unqualified target ball circles in the first circular set Y; otherwise, record the target ball circle with the current radius Rt0 as a qualified target ball circle, retain the qualified target ball circles, and generate the second circular set Y0. S6. Preprocess the qualified target sphere circles to obtain the target sphere point cloud; S7. Fit the target sphere model to the target sphere point cloud and determine whether the fitting result of the target sphere model is greater than the residual threshold. If so, remove the unqualified target sphere model; otherwise, retain the qualified target sphere model and use it as the target sphere recognition result.

2. The automatic identification method for laser scanning point cloud target spheres according to claim 1, characterized in that, The specific process of converting 3D laser scanning point cloud data into a depth map includes: S21. Select the location of the station center as the reference point, and calculate the distance from each point in the 3D laser scanning point cloud data to the station center; S22. Calculate the horizontal and vertical angles from each point in the 3D laser scanning point cloud data to the center of the station; S23. Combine the distance, horizontal angle, and pitch angle of each point in the 3D laser scanning point cloud data to the center of the station to obtain the depth information of each point in the 3D laser scanning point cloud data, and generate a depth map P.

3. The automatic identification method for laser scanning point cloud target spheres according to claim 2, characterized in that, The specific process of performing edge enhancement filtering on the depth map includes: S31. Perform a Laplacian transform on the depth map to generate an image with enhanced edges; S32. Apply Gaussian filtering to the edge-enhanced image, and then use the Canny operator to perform edge detection on the Gaussian-filtered image to generate a filtered image.

4. The automatic identification method for laser scanning point cloud target spheres according to claim 3, characterized in that, The specific process of identifying and extracting circles in the filtered image includes: performing circle identification and extraction in the filtered image through Hough transform to obtain a first set of circles Y, and denoteing the circles in the first set of circles Y as target sphere circles, with the radius of the circle being the radius Rt0 of the target sphere circle.

5. The automatic identification method for laser scanning point cloud target spheres according to claim 4, characterized in that, The pre-processing of qualified target ball circles includes the following steps: S61. In the second circular set Y0, the coordinates of the center pixel point inside each target ball circle in the second circular set Y0 are converted into the three-dimensional coordinates Pmc of the center of the target ball; S62. The three-dimensional coordinates Pmc of the center of the target sphere described in S61 are denoted as the target sphere point cloud, and the target sphere point cloud is used to form a target sphere point cloud set PC.

6. The automatic identification method for laser scanning point cloud target spheres according to claim 5, characterized in that, The process of fitting a target sphere model to the target sphere point cloud and determining whether the fitting result of the target sphere model is greater than the residual threshold includes: S71. Construct a target sphere model based on the three-dimensional coordinates Pmc of the center of each target sphere in the target sphere point cloud set PC and the radius R of each target sphere; S72. Calculate the distance residual rms from each point on the target ball point cloud in the target ball point cloud set PC to the corresponding target ball model surface, and denote the distance residual rms as the fitting result; S73. Determine whether the distance residual rms is greater than the residual threshold. If so, the fitting result of the target ball model is greater than the residual threshold, and remove the unqualified target ball model from the second circular set Y0; otherwise, leave the qualified target ball model in the second circular set Y0 and use it as the identification result of the target ball.

7. The automatic identification method for laser scanning point cloud target spheres according to any one of claims 1-6, characterized in that, The three-dimensional laser scanning point cloud data is obtained by scanning the target with a three-dimensional laser scanner.

8. An automatic identification system for laser-scanned point cloud target spheres, said system being implemented based on the automatic identification method for laser-scanned point cloud target spheres according to any one of claims 1-7, characterized in that, The system includes: The acquisition module is used to acquire the three-dimensional laser scanning point cloud data of the target; The conversion module is used to convert 3D laser scanning point cloud data into depth maps; The enhancement filtering module is used to perform edge enhancement filtering on the depth map to obtain a filtered image. The identification and extraction module is used to identify and extract circles in the filtered image, and the circles are referred to as target spheres. A filtering module is used to filter the target ball circles to obtain qualified target ball circles; The preprocessing module is used to preprocess qualified target sphere circles to obtain target sphere point clouds; The judgment module is used to fit the target sphere model to the target sphere point cloud and determine whether the fitting result of the target sphere model is greater than the residual threshold. If so, the unqualified target sphere model is removed; otherwise, the qualified target sphere model is retained and used as the recognition result of the target sphere.

9. A computer device, characterized in that, include: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction that causes the processor to perform the operation of the automatic identification method for laser scanning point cloud target spheres as described in any one of claims 1-7.

Images