Point cloud automatic classification-based noise post-processing method and device, equipment and medium

Abstract

The present disclosure provides a noise point post-processing method and device based on point cloud automatic classification, equipment and medium. The present disclosure first classifies the input point cloud data, then establishes a KD tree for the point cloud data between adjacent towers, searches for neighboring points and counts the average distance, target distance mean and distance standard deviation of each point; then according to the category of the point cloud data classification, different preset standard deviation multipliers are determined, and the distance threshold of each category is further determined; then the points with an average distance exceeding the corresponding distance threshold are screened as potential noise points; then the lowest height value of the conductor is taken as the demarcation to divide the point cloud data into two regions above and below the conductor for separate processing to determine the final target noise point. The problems of low classification efficiency and high dependence on training data of point cloud data noise points in the power scene are solved, the accuracy of noise point classification results and the automation degree of data production are improved, and the demand for computer hardware and a large amount of manual annotation is reduced.

Description

Technical Field

[0001] This disclosure relates to the field of power data processing, and in particular to a noise post-processing method, apparatus, equipment, and medium based on point cloud automatic classification. Background Technology

[0002] When processing point cloud data in power scenarios, although many noise extraction algorithms have been proposed, the complexity, diversity and other interference of power scenarios make it difficult to achieve good noise classification results using a single rule or pure deep learning.

[0003] Existing noise classification schemes all have limitations and drawbacks. For example, clustering-based noise classification methods use manually set cluster radii and minimum number of points per cluster to filter out outliers and classify them as noise. This method suffers from significant performance issues. When the point cloud is large, the clustering algorithm consumes a lot of memory, places high demands on computer hardware during parallel processing, and is quite time-consuming. Furthermore, this method requires extensive human experience and parameter tuning, making it difficult to apply to complex and varied real-world scenarios. The rules cannot cover all possible situations, leading to missed detections and false positives. Another example is machine learning or deep learning-based classification methods, which directly learn features and model from the raw point cloud data for automatic noise classification. This method is highly dependent on training data, requiring a large amount of well-labeled training data, and has poor robustness to noisy point clouds. Summary of the Invention

[0004] This disclosure provides at least one method, apparatus, device, and medium for noise post-processing based on point cloud automatic classification, in order to solve at least one of the above-mentioned technical defects.

[0005] According to one aspect of this disclosure, a noise post-processing method based on automatic point cloud classification is provided, comprising:

[0006] The raw point cloud data obtained by scanning the target power line area with a 3D lidar is acquired, the raw point cloud data is classified and processed, and point cloud data of each category is extracted; the categories include conductor category and noise category.

[0007] The minimum height value of the conductor is determined based on the point cloud data corresponding to the conductor category;

[0008] Extract the point cloud data between every two adjacent towers from the point cloud data of the above categories to obtain multiple tower-tower point cloud arrays;

[0009] Perform the following operations on the point cloud array for each tower:

[0010] Construct a KD tree for the point cloud array between the towers;

[0011] Iterate through each point in the point cloud array between the towers. For each point, use the KD tree to search for the N nearest points to that point and calculate the average distance between the N nearest points and the point; where N is a positive integer.

[0012] Calculate the mean of the average distances corresponding to all points in the point cloud array between the towers to obtain the mean target distance; and calculate the standard deviation of the distance based on the average distances corresponding to all points.

[0013] Based on the mean target distance, the standard deviation of the distance, and the preset standard deviation multiplier corresponding to each category, determine the distance threshold corresponding to each category;

[0014] Iterate through each point in the point cloud array between the towers. For each point, if the average distance to that point is greater than the distance threshold of the category to which that point belongs, then that point is considered a potential noise point.

[0015] Traverse all potential noise points. If the height value of a potential noise point is less than the minimum height value of the conductor, use the KD tree to search for points that are less than a preset distance from the potential noise point. If the number of points found is zero, then the potential noise point is determined to be the target noise point. If the height value of a potential noise point is greater than the minimum height value of the conductor, then obtain the N points that are closest to the potential noise point. If the category of the N points is all noise points, then the potential noise point is determined to be the target noise point.

[0016] In one possible implementation, the category further includes at least one of the following:

[0017] Ground wire category; crossover wire category; pole / tower category; insulator drain wire category; vegetation category; ground category.

[0018] In one possible implementation, the preset standard deviation multiplier corresponding to the tower category is greater than the preset standard deviation multiplier corresponding to the vegetation category and the ground category; the preset standard deviation multiplier corresponding to the conductor category, ground wire category, crossover wire category, and insulator drain wire category is greater than the preset standard deviation multiplier corresponding to the vegetation category and the ground category.

[0019] In one possible implementation, determining the distance threshold corresponding to each category based on the target distance mean, the distance standard deviation, and the preset standard deviation multiplier for each category includes:

[0020] For each category, the preset standard deviation multiplier corresponding to the category is multiplied by the distance standard deviation, and the resulting value is added to the mean of the target distance to obtain the distance threshold corresponding to the category.

[0021] In one possible implementation, determining the minimum height value of the conductor based on the point cloud data corresponding to the conductor category includes:

[0022] Sort the points in the point cloud data corresponding to the traverse category according to the elevation values ​​from low to high to obtain the target point cloud sequence;

[0023] According to the preset elevation values, the target point cloud sequence is segmented to obtain target point cloud arrays corresponding to multiple elevation segments;

[0024] The minimum height of the traverse is obtained by calculating the average elevation values ​​of all points in the target point cloud array corresponding to the first M lowest elevation segments; where M is a positive integer.

[0025] In one possible implementation, the classification process of the raw point cloud data includes:

[0026] The raw point cloud data between every two adjacent towers is divided to obtain multiple sets of raw point cloud arrays;

[0027] Deep learning was used to classify the original point cloud arrays for each group.

[0028] In one possible implementation, obtaining the N points closest to the potential noise point includes:

[0029] The KD tree search finds the N points closest to the potential noise.

[0030] According to another aspect of this disclosure, a noise post-processing apparatus based on automatic point cloud classification is provided, comprising:

[0031] The point cloud acquisition and processing module is used to acquire raw point cloud data obtained by scanning the target power line area with a 3D lidar, classify the raw point cloud data, and extract point cloud data of each category; the categories include conductor category and noise category.

[0032] The minimum conductive height determination module is used to determine the minimum height value of the conductor based on the point cloud data corresponding to the conductor type.

[0033] The point cloud segmentation module is used to extract the point cloud data between every two adjacent towers from the point cloud data of the various categories, and obtain multiple tower-tower point cloud arrays.

[0034] The noise post-processing module performs the following operations on the point cloud array for each tower:

[0035] Construct a KD tree for the point cloud array between the towers;

[0036] Iterate through each point in the point cloud array between the towers. For each point, use the KD tree to search for the N nearest points to that point and calculate the average distance between the N nearest points and the point; where N is a positive integer.

[0037] Calculate the mean of the average distances corresponding to all points in the point cloud array between the towers to obtain the mean target distance; and calculate the standard deviation of the distance based on the average distances corresponding to all points.

[0038] Based on the mean target distance, the standard deviation of the distance, and the preset standard deviation multiplier corresponding to each category, determine the distance threshold corresponding to each category;

[0039] Iterate through each point in the point cloud array between the towers. For each point, if the average distance to that point is greater than the distance threshold of the category to which that point belongs, then that point is considered a potential noise point.

[0040] Traverse all potential noise points. If the height value of a potential noise point is less than the minimum height value of the conductor, use the KD tree to search for points that are less than a preset distance from the potential noise point. If the number of points found is zero, then the potential noise point is determined to be the target noise point. If the height value of a potential noise point is greater than the minimum height value of the conductor, then obtain the N points that are closest to the potential noise point. If the category of the N points is all noise points, then the potential noise point is determined to be the target noise point.

[0041] According to another aspect of this disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory, wherein the processor, when executing the computer program, implements the method described in any of the preceding claims.

[0042] According to another aspect of this disclosure, a computer-readable storage medium is provided, wherein a computer program is stored therein, and the computer program, when executed by a processor, implements the method described in any of the preceding claims.

[0043] This disclosed noise post-processing method based on automatic point cloud classification first classifies the input point cloud data. Then, a kd-tree is built for the point cloud data between adjacent towers to search for neighboring points and calculate the average distance, target distance mean, and distance standard deviation for each point. Next, based on the classification category of the point cloud data, different preset standard deviation multipliers are determined, and distance thresholds for each category are further defined. Points with an average distance exceeding the corresponding distance threshold are then selected as potential noise points. Finally, the area is divided into two regions, above and below the conductor, using the lowest conductor height as a boundary for separate processing to determine the final target noise points. This method solves the problems of low efficiency in point cloud data noise classification and high dependence on training data in power scenarios, improves the accuracy of noise classification results and the automation of data production, and reduces the need for computer hardware and extensive manual annotation.

[0044] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0045] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0046] Figure 1 This is a flowchart of a noise post-processing method based on point cloud automatic classification according to this disclosure;

[0047] Figure 2 This is a structural schematic diagram of a noise post-processing device based on point cloud automatic classification according to the present disclosure;

[0048] Figure 3 This is a schematic diagram of the structure of an electronic device according to the present disclosure. Detailed Implementation

[0049] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0050] This disclosure addresses the shortcomings of current point cloud data noise classification methods, such as low efficiency, low accuracy, high labor costs, and poor robustness. It provides a noise post-processing method, apparatus, device, and medium based on automatic point cloud classification. This disclosure solves the problems of low efficiency and high dependence on training data in point cloud data noise classification in power scenarios, improves the accuracy of noise classification results, the degree of automation in data production, and the adaptability to different scenarios, and reduces the need for computer hardware and a large amount of manual annotation.

[0051] The technical solution of this disclosure will be described below through specific embodiments.

[0052] like Figure 1 The diagram shown is a flowchart of the noise post-processing method based on automatic point cloud classification in this embodiment. The execution subject of this embodiment is a computing device or component with data processing capabilities. Specifically, the method of this embodiment may include the following steps:

[0053] S110. Obtain the raw point cloud data obtained by scanning the target power line area with a three-dimensional lidar, classify the raw point cloud data, and extract point cloud data of each category; the categories include conductor category and noise category.

[0054] The collected raw point cloud data includes features such as power lines, poles, ground, buildings, and vegetation. Therefore, the above classification processing of the raw point cloud data can yield the following categories of point cloud data: ground wire category; crossover line category; pole and tower category; insulator drain line category; vegetation category; ground category; and building category.

[0055] The classification and processing of the raw point cloud data can be achieved using the following steps:

[0056] The raw point cloud data between every two adjacent towers is segmented to obtain multiple sets of raw point cloud arrays; deep learning is used to classify each set of raw point cloud arrays.

[0057] The original point cloud data between every two adjacent towers is divided into multiple sets of point cloud data, which reduces the amount of data and facilitates subsequent separate processing.

[0058] In addition, each set of original point cloud arrays can be automatically classified using methods such as clustering. Point cloud data classification helps to filter out useless points and obtain the required point cloud data. In this invention, the point cloud data after the initial classification is extracted for post-processing. For example, in step S120, the minimum height value of the conductor is determined using the point cloud data of the conductor category.

[0059] S120. Determine the minimum height value of the conductor based on the point cloud data corresponding to the conductor category.

[0060] It should be noted that step S120 does not necessarily have to be executed before step S130. This disclosure does not limit the execution order of steps by their numbers.

[0061] Subsequently, the minimum height value of the conductor will be used to divide the entire area into two regions: the area above the conductor and the area below the conductor, and noise will be classified and processed separately for each region.

[0062] S130. Extract the point cloud data between every two adjacent towers from the point cloud data of the various categories to obtain multiple tower-tower point cloud arrays.

[0063] Processing the point cloud data between each pair of adjacent towers separately can reduce the amount of data processed per cycle and improve processing efficiency.

[0064] S140. Perform noise classification for each point cloud array between towers to determine the target noise in each point cloud array between towers:

[0065] Step 1: Construct a KD tree for the point cloud array between towers.

[0066] KD trees are used for subsequent searches of neighboring points.

[0067] Step 2: Traverse each point in the point cloud array between the towers. For each point, use the KD tree to search for the N nearest points to that point and calculate the average distance between the N nearest points and the point; where N is a positive integer.

[0068] The value of N here can be determined based on the needs of the actual scenario; for example, N can be set to 5.

[0069] Step 3: Calculate the mean of the average distances corresponding to all points in the point cloud array between the towers to obtain the mean target distance; and calculate the standard deviation of the distance based on the average distances corresponding to all points.

[0070] Step 4: Determine the distance threshold corresponding to each category based on the mean of the target distance, the standard deviation of the distance, and the preset standard deviation multiplier corresponding to each category.

[0071] Different preset standard deviation multipliers are used for different categories of point cloud data. In order to avoid important targets such as power lines and towers being misclassified as noise, the preset standard deviation multipliers for these categories are set to be larger. Meanwhile, targets such as vegetation and ground contain a lot of noise, so the preset standard deviation multipliers for these categories are set to be smaller.

[0072] Specifically, the preset standard deviation multiplier corresponding to the tower category is greater than the preset standard deviation multiplier corresponding to the vegetation category and the ground category; the preset standard deviation multiplier corresponding to the conductor category, ground wire category, crossover wire category, and insulator drain wire category is greater than the preset standard deviation multiplier corresponding to the vegetation category and the ground category.

[0073] Step 5: Traverse each point in the point cloud array between the towers. For each point, if the average distance to that point is greater than the distance threshold of the category to which that point belongs, then that point is considered a potential noise point.

[0074] Step Six: Traverse all potential noise points. If the height of a potential noise point is less than the minimum height of the conductor, use the KD-tree to search for points that are less than a preset distance from the potential noise point. If the number of searched points is zero, the potential noise point is determined to be a target noise point. Conversely, if the number of searched points is not zero, that is, other points are searched, the potential noise point is determined not to be a target noise point. If the height of a potential noise point is greater than the minimum height of the conductor, obtain the N points closest to the potential noise point. If all N points belong to the noise category, the potential noise point is determined to be a target noise point. That is, if none of the N points belong to any category other than the noise category, the potential noise point is determined to be a target noise point.

[0075] Here, we can directly use the N points found in step two, or we can use a KD tree to search again for the N points closest to the potential noise point.

[0076] In some embodiments, the distance threshold corresponding to each category is determined based on the mean target distance, the standard deviation of the distance, and the preset standard deviation multiplier corresponding to each category. This can be achieved using the following steps:

[0077] For each category, the preset standard deviation multiplier corresponding to the category is multiplied by the distance standard deviation, and the resulting value is added to the mean of the target distance to obtain the distance threshold corresponding to the category.

[0078] In some embodiments, the minimum height value of the conductor is determined based on the point cloud data corresponding to the conductor category. This can be achieved through the following steps:

[0079] Step 1: Sort the points in the point cloud data corresponding to the traverse category according to the elevation values ​​from low to high to obtain the target point cloud sequence.

[0080] Step 2: According to the preset elevation values, the target point cloud sequence is segmented to obtain target point cloud arrays corresponding to multiple elevation segments.

[0081] Step 3: Calculate the average elevation value of all points in the target point cloud array corresponding to the first M lowest elevation segments to obtain the minimum height value of the traverse; where M is a positive integer.

[0082] Specifically, the point cloud data corresponding to the traverse category in step S120 is extracted, and the point cloud data is sorted from low to high elevation. Starting from the lowest point, all points are segmented in 0.5-meter elevation increments (i.e., preset elevation values). The points in the first three (M) lowest elevation segments are extracted, and the average elevation of these points is calculated. This average value is used as the dividing standard, i.e., the minimum elevation value of the traverse. The purpose of using the minimum elevation value of the traverse to classify noise points by region is to overcome the following problem: there are a large number of ground points and vegetation points with lower elevation values. Some points, although they are outliers and are located in the vegetation layer or below the ground, are still classified as vegetation points or ground points. To avoid classifying these points as noise points, the minimum elevation value of the traverse is used as the dividing line for separate processing.

[0083] The solutions described in the above embodiments improve the accuracy of automatic noise classification in power scenarios and more effectively support other business needs based on the automated classification results. Furthermore, compared to methods such as clustering denoising, the solutions in the above embodiments consume less memory, have higher execution efficiency, and lower requirements for computer hardware. In addition, the solutions in the above embodiments do not rely on large amounts of training data, do not require significant manpower to label good samples, and exhibit high robustness to noise in various scenarios.

[0084] Based on the same inventive concept, this disclosure provides a noise post-processing device based on automatic point cloud classification. The steps performed by the components of this device are the same as or similar to those of the method described above; therefore, similar aspects will not be repeated. Figure 2 As shown, the noise post-processing device based on point cloud automatic classification in this embodiment includes:

[0085] The point cloud acquisition and processing module 210 is used to acquire the raw point cloud data obtained by scanning the target power line area by the three-dimensional lidar, classify the raw point cloud data, and extract point cloud data of each category; the categories include conductor category and noise category.

[0086] The minimum conductive height determination module 220 is used to determine the minimum height value of the conductor based on the point cloud data corresponding to the conductor type.

[0087] The point cloud segmentation module 230 is used to extract point cloud data between every two adjacent towers from the point cloud data of the various categories, and obtain multiple tower-tower point cloud arrays.

[0088] Noise post-processing module 240 is used to perform the following operations on the point cloud array between each tower:

[0089] Construct a KD tree for the point cloud array between the towers;

[0090] Iterate through each point in the point cloud array between the towers. For each point, use the KD tree to search for the N nearest points to that point and calculate the average distance between the N nearest points and the point; where N is a positive integer.

[0091] Calculate the mean of the average distances corresponding to all points in the point cloud array between the towers to obtain the mean target distance; and calculate the standard deviation of the distance based on the average distances corresponding to all points.

[0092] Based on the mean target distance, the standard deviation of the distance, and the preset standard deviation multiplier corresponding to each category, determine the distance threshold corresponding to each category;

[0093] Iterate through each point in the point cloud array between the towers. For each point, if the average distance to that point is greater than the distance threshold of the category to which that point belongs, then that point is considered a potential noise point.

[0094] Traverse all potential noise points. If the height value of a potential noise point is less than the minimum height value of the conductor, use the KD tree to search for points that are less than a preset distance from the potential noise point. If the number of points found is zero, then the potential noise point is determined to be the target noise point. If the height value of a potential noise point is greater than the minimum height value of the conductor, then obtain the N points that are closest to the potential noise point. If the category of the N points is all noise points, then the potential noise point is determined to be the target noise point.

[0095] In some embodiments, the category of point cloud data also includes at least one of the following:

[0096] Ground wire category; crossover wire category; pole / tower category; insulator drain wire category; vegetation category; ground category.

[0097] In some embodiments, the preset standard deviation multiplier corresponding to the tower category is greater than the preset standard deviation multiplier corresponding to the vegetation category and the ground category; the preset standard deviation multiplier corresponding to the conductor category, ground wire category, crossover wire category, and insulator drain wire category is greater than the preset standard deviation multiplier corresponding to the vegetation category and the ground category.

[0098] In some embodiments, when the noise post-processing module 240 determines the distance threshold corresponding to each category based on the target distance mean, the distance standard deviation, and the preset standard deviation multiplier corresponding to each category, it is specifically used for:

[0099] For each category, the preset standard deviation multiplier corresponding to the category is multiplied by the distance standard deviation, and the resulting value is added to the mean of the target distance to obtain the distance threshold corresponding to the category.

[0100] In some embodiments, when the minimum conductive height determination module 220 determines the minimum conductor height value based on the point cloud data corresponding to the conductor type, it is specifically used for:

[0101] Sort the points in the point cloud data corresponding to the traverse category according to the elevation values ​​from low to high to obtain the target point cloud sequence;

[0102] According to the preset elevation values, the target point cloud sequence is segmented to obtain target point cloud arrays corresponding to multiple elevation segments;

[0103] The minimum height of the traverse is obtained by calculating the average elevation values ​​of all points in the target point cloud array corresponding to the first M lowest elevation segments; where M is a positive integer.

[0104] In some embodiments, when the point cloud acquisition and processing module 210 performs classification processing on the raw point cloud data, it is specifically used for:

[0105] The raw point cloud data between every two adjacent towers is divided to obtain multiple sets of raw point cloud arrays;

[0106] Deep learning was used to classify the original point cloud arrays for each group.

[0107] In some embodiments, when the noise post-processing module 240 obtains the N points closest to the potential noise, it is specifically used for:

[0108] The KD tree search finds the N points closest to the potential noise.

[0109] According to embodiments of this disclosure, this disclosure also provides an electronic device and a computer-readable storage medium.

[0110] Figure 3 A schematic block diagram of an example electronic device 300 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0111] like Figure 3As shown, device 300 includes a computing unit 310, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 320 or a computer program loaded from storage unit 380 into random access memory (RAM) 330. The RAM 330 may also store various programs and data required for the operation of device 300. The computing unit 310, ROM 320, and RAM 330 are interconnected via bus 340. Input / output (I / O) interface 350 is also connected to bus 340.

[0112] Multiple components in device 300 are connected to I / O interface 350, including: input unit 360, such as keyboard, mouse, etc.; output unit 370, such as various types of monitors, speakers, etc.; storage unit 380, such as disk, optical disk, etc.; and communication unit 390, such as network card, modem, wireless transceiver, etc. Communication unit 390 allows device 300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0113] The computing unit 310 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 310 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 310 performs the various methods and processes described above. For example, in some embodiments, any of the methods described above can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 380. In some embodiments, part or all of the computer program can be loaded and / or installed on device 300 via ROM 320 and / or communication unit 390. When the computer program is loaded into RAM 330 and executed by the computing unit 310, one or more steps of any of the methods described above can be performed. Alternatively, in other embodiments, the computing unit 310 can be configured to perform any of the methods described above by any other suitable means (e.g., by means of firmware).

[0114] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0115] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0116] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0117] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0118] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0119] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0120] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0121] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A noise post-processing method based on automatic point cloud classification, characterized in that, include: The raw point cloud data obtained by scanning the target power line area with a 3D lidar is acquired, the raw point cloud data is classified and processed, and point cloud data of each category is extracted; the categories include conductor category and noise category. The minimum height value of the conductor is determined based on the point cloud data corresponding to the conductor category; Extract the point cloud data between every two adjacent towers from the point cloud data of the above categories to obtain multiple tower-tower point cloud arrays; Perform the following operations on the point cloud array for each tower: Construct a KD tree for the point cloud array between the towers; Iterate through each point in the point cloud array between the towers. For each point, use the KD tree to search for the N nearest points to that point and calculate the average distance between the N nearest points and the point; where N is a positive integer. Calculate the mean of the average distances corresponding to all points in the point cloud array between the towers to obtain the mean target distance; and calculate the standard deviation of the distance based on the average distances corresponding to all points. Based on the mean target distance, the standard deviation of the distance, and the preset standard deviation multiplier corresponding to each category, determine the distance threshold corresponding to each category; Iterate through each point in the point cloud array between the towers. For each point, if the average distance to that point is greater than the distance threshold of the category to which that point is located, then that point is considered a potential noise point. Traverse all potential noise points. If the height value of a potential noise point is less than the minimum height value of the conductor, use the KD tree to search for points that are less than a preset distance from the potential noise point. If the number of points found is zero, then the potential noise point is determined to be the target noise point. If the height value of a potential noise point is greater than the minimum height value of the conductor, then obtain the N points that are closest to the potential noise point. If the category of the N points is all noise points, then the potential noise point is determined to be the target noise point.

2. The method according to claim 1, characterized in that, The category also includes at least one of the following: Ground wire category; crossover wire category; pole / tower category; insulator drain wire category; vegetation category; ground category.

3. The method according to claim 2, characterized in that, The preset standard deviation multiplier corresponding to the tower category is greater than the preset standard deviation multiplier corresponding to the vegetation category and the ground category; the preset standard deviation multiplier corresponding to the conductor category, ground wire category, crossover wire category, and insulator drain wire category is greater than the preset standard deviation multiplier corresponding to the vegetation category and the ground category.

4. The method according to claim 1, characterized in that, The step of determining the distance threshold corresponding to each category based on the target distance mean, the distance standard deviation, and the preset standard deviation multiplier corresponding to each category includes: For each category, the preset standard deviation multiplier corresponding to the category is multiplied by the distance standard deviation, and the resulting value is added to the mean of the target distance to obtain the distance threshold corresponding to the category.

5. The method according to claim 1, characterized in that, The step of determining the minimum height value of the conductor based on the point cloud data corresponding to the conductor category includes: Sort the points in the point cloud data corresponding to the traverse category according to the elevation values ​​from low to high to obtain the target point cloud sequence; According to the preset elevation values, the target point cloud sequence is segmented to obtain target point cloud arrays corresponding to multiple elevation segments; The minimum height of the traverse is obtained by calculating the average elevation values ​​of all points in the target point cloud array corresponding to the first M lowest elevation segments; where M is a positive integer.

6. The method according to claim 1, characterized in that, The classification process for the raw point cloud data includes: The raw point cloud data between every two adjacent towers is divided to obtain multiple sets of raw point cloud arrays; Deep learning was used to classify the original point cloud arrays for each group.

7. The method according to claim 1, characterized in that, The process of obtaining the N points closest to the potential noise includes: The KD tree search finds the N points closest to the potential noise.

8. A noise post-processing device based on automatic point cloud classification, characterized in that, include: The point cloud acquisition and processing module is used to acquire raw point cloud data obtained by scanning the target power line area with a 3D lidar, classify the raw point cloud data, and extract point cloud data of each category; the categories include conductor category and noise category. The minimum conductive height determination module is used to determine the minimum height value of the conductor based on the point cloud data corresponding to the conductor type. The point cloud segmentation module is used to extract the point cloud data between every two adjacent towers from the point cloud data of the various categories, and obtain multiple tower-tower point cloud arrays. The noise post-processing module performs the following operations on the point cloud array for each tower: Construct a KD tree for the point cloud array between the towers; Iterate through each point in the point cloud array between the towers. For each point, use the KD tree to search for the N nearest points to that point and calculate the average distance between the N nearest points and the point; where N is a positive integer. Calculate the mean of the average distances corresponding to all points in the point cloud array between the towers to obtain the mean target distance; and calculate the standard deviation of the distance based on the average distances corresponding to all points. Based on the mean target distance, the standard deviation of the distance, and the preset standard deviation multiplier corresponding to each category, determine the distance threshold corresponding to each category; Iterate through each point in the point cloud array between the towers. For each point, if the average distance to that point is greater than the distance threshold of the category to which that point is located, then that point is considered a potential noise point. Traverse all potential noise points. If the height value of a potential noise point is less than the minimum height value of the conductor, use the KD tree to search for points that are less than a preset distance from the potential noise point. If the number of points found is zero, then the potential noise point is determined to be the target noise point. If the height value of a potential noise point is greater than the minimum height value of the conductor, then obtain the N points that are closest to the potential noise point. If the category of the N points is all noise points, then the potential noise point is determined to be the target noise point.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory, wherein the processor, when executing the computer program, implements the method of any one of claims 1-7.

10. A computer-readable storage medium storing a computer program that, when executed by a processor, implements the method of any one of claims 1-7.

Images