Underground cable and pipe trench surveying and mapping and collecting and recording method based on artificial intelligence

Abstract

The invention discloses an underground cable and pipe trench surveying and mapping and collecting and recording method based on artificial intelligence, and belongs to the technical field of artificial intelligence, and the method comprises the steps: scanning the ground and underground of a cable well, and obtaining the first transmission information of the cable well; transmitting the first transmission information to a data platform based on the data transmission mode and storing the first transmission information in a database; extracting three-dimensional point cloud data from the first transmission information, performing preliminary cutting and denoising to obtain standard point cloud data, obtaining spatial information of the cable well based on the standard point cloud data, and performing model reconstruction on the cable well to obtain a three-dimensional model of the cable well; extracting image data from the first transmission information, identifying and marking key features in the image data, training a machine learning model by taking the marked key features as a training set, evaluating the trained machine learning model, and marking predicted and generated underground detection data on a three-dimensional model. According to the invention, the identification efficiency of the underground defect of the cable well is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence, and particularly relates to a surveying, mapping, collecting and recording method for underground cables and cable trenches based on artificial intelligence. Background Art

[0002] With the development of cities and the improvement of the cable rate, power cables, as the lifeline of cities, their management level directly affects power supply reliability and urban public safety. Power cable management systems based on geographic information systems (GIS) are increasingly being applied. However, most systems are based on two-dimensional coordinates and have limitations in spatial representation and data analysis. To improve the efficiency and accuracy of cable management, researchers at home and abroad have proposed various new cable management system models, including those based on three-dimensional simulation technology, ground penetrating radar technology, three-layer GIS technology, RFID technology, etc.

[0003] Existing ones, such as the Chinese patent application with the publication number CN112215197A, propose an underground cable fault warning method based on artificial intelligence. The monitoring and warning master station sends an inquiry signal to the cable parameter detection unit; after receiving the inquiry signal, the cable parameter detection unit starts cable parameter collection and sends the detection value to the monitoring and warning master station; after receiving the detection value, the monitoring and warning master station runs an early warning algorithm of artificial intelligence based on a convolutional neural network to judge the state of the underground cable and give an early warning of cable faults.

[0004] The above-mentioned existing technology uses an early warning algorithm of artificial intelligence based on a convolutional neural network to give an early warning of cable faults. However, a three-dimensional model of the underground cable is not constructed, and there are also certain deficiencies in the defect recognition accuracy. Therefore, a surveying, mapping, collecting and recording method for underground cables and cable trenches based on artificial intelligence is needed to improve the management efficiency of underground cables. Summary of the Invention

[0005] To solve the above problems, the present invention provides a surveying, mapping, collecting and recording method for underground cables and cable trenches based on artificial intelligence to solve the problems in the prior art.

[0006] To achieve the above invention purpose, the present invention proposes a surveying, mapping, collecting and recording method for underground cables and cable trenches based on artificial intelligence, including: Step S1: Scanning the ground and underground of the cable well based on surveying equipment to obtain the first transmission information of the cable well, where the first transmission information includes three-dimensional point cloud data and image data; Step S2: Setting a data transmission method, and transmitting the first transmission information to a data platform based on the data transmission method and storing it in a database; Step S3: Extract the 3D point cloud data from the first transmission information in the database, perform preliminary cropping on the 3D point cloud data, denoise the cropped 3D point cloud data based on the RANSAC algorithm to obtain standard point cloud data, obtain the spatial information of the cable shaft based on the standard point cloud data, and use 3D point cloud reconstruction technology to reconstruct the model of the cable shaft to obtain the 3D model of the cable shaft; Step S4: Extract the image data from the first transmission information in the database, identify the key features in the image data, label the key features, establish a machine learning model based on the YOLOv8 algorithm, the machine learning model is used to detect the underground defects of the cable shaft and identify the occupancy of the pipe openings, set the learning rate of the model, use the labeled key features as the training set to train the machine learning model, evaluate the trained machine learning model to obtain the accuracy of the model. If the accuracy is greater than the first threshold, use the predicted underground defect data and the identification data of whether the pipe opening is occupied as the underground detection data, and label the underground detection data on the 3D model.

[0007] Further, transmitting the first transmission information to the data platform and storing it in the database includes the following steps: Encapsulate the first transmission information into a data packet, add header information before the data packet, the header information includes the port number of the communication protocol used by the surveying and mapping device, define the data packet with the added header information as the second transmission data, set up a gateway device, and transmit the second transmission data to the gateway device. The gateway device includes a data connection unit and a data queuing unit. The data connection unit removes the header information to extract the data packet, generates an encryption key, encrypts the data packet, assigns label information to the encrypted data packet, and defines it as the third transmission data; The data queuing unit sets up a queue list, there is a corresponding relationship between multiple queues and label information in the queue list. The data queuing unit stores the third transmission data in the corresponding queue based on the queue list, and converts the third transmission data into the fourth transmission data. The fourth transmission data is data that can be received by the data cloud platform after protocol conversion and format conversion. The data cloud platform obtains the fourth transmission data from the data queuing unit, decrypts the fourth transmission data using the key corresponding to the encrypted data, and stores the decrypted fourth transmission data in the database.

[0008] Further, obtaining the standard point cloud data includes the following steps: Step S31: randomly selecting a portion of data from the three-dimensional point cloud data as first sample data based on the RANSAC algorithm, fitting and generating a geometric model of the bottom or side of the cable well based on the first sample data, calculating a vertical distance from each three-dimensional point cloud data in the first sample data to the geometric model, and if the vertical distance is less than a second threshold, defining the three-dimensional point cloud data as first-category point cloud data, otherwise defining it as second-category point cloud data, respectively calculating the number of first-category point cloud data and second-category point cloud data in the first sample data, defining them as a first number and a second number, and eliminating the second-category point cloud data in the first sample data; Step S32: Based on the RANSAC algorithm, randomly select a part of the data from the remaining three-dimensional point cloud data as the second sample data, repeat step S31, and calculate the ratio of the second number to the first number after each iteration. If the ratio is less than the third threshold, the denoising process of the RANSAC algorithm ends, and the remaining first-category point cloud data is defined as standard point cloud data.

[0009] Furthermore, obtaining the spatial information of the cable well includes the following steps: Obtain the maximum coordinate value and the minimum coordinate value of the standard point cloud data along the three coordinate axes of X, Y and Z, calculate the difference between the maximum coordinate value and the minimum coordinate value on each coordinate axis, define them as the first length, the first width and the first height respectively, construct a bounding box of the standard point cloud data based on the first length, the first width and the first height, and obtain the spatial information of the cable well based on the size of the bounding box, wherein the spatial information includes the length, width and height of the cable well.

[0010] Furthermore, obtaining the three-dimensional model of the cable well comprises the following steps: The standard point cloud data is mesh reconstructed, the standard point cloud data is converted into a mesh model of multiple faces, the mesh model is smoothed, and a three-dimensional model of the cable well is generated.

[0011] Furthermore, building a machine learning model based on the YOLOv8 algorithm includes the following steps: The machine learning model classifies the downhole defect information to obtain multiple defect categories, defines the defects predicted by the machine learning model as the first label, and defines the actual downhole defects as the second label. The loss function of the YOLOv8 algorithm is set, and the difference between the first label and the second label is calculated based on the loss function. The loss function is: ,in, is the predicted probability of the machine learning model for each of the defect categories, is a coefficient for balancing the weights of positive and negative samples. Positive class samples refer to cases with defects, and negative class samples refer to cases without defects. is a parameter for adjusting the sample weights. is the difference value. Based on the difference value, the model parameters of the machine learning model are updated. When the difference value is less than the fourth threshold, the construction of the machine learning model is completed.

[0012] Furthermore, the key features include the occupancy of the cross-section and pipe orifices, and the defect information underground in the cable trench.

[0013] Furthermore, setting the learning rate of the model includes the following steps: Continuously adjust the learning rate during the model training process based on the first formula , and the first formula is: , where is the initial learning rate, is the preset decay rate, is the number of training times, and the machine learning model is trained based on the learning rate.

[0014] Furthermore, evaluating the trained machine learning model to obtain the accuracy of the model includes the following steps: Evaluate the machine learning model using the test set, and calculate the accuracy based on the second formula , and the second formula is: , where is the number of samples correctly predicted as positive class by the model, is the number of samples correctly predicted as negative class by the model, is the number of samples wrongly predicted as positive class by the model, is the number of samples wrongly predicted as negative class by the model.

[0015] Furthermore, after detecting the defect information of the cable trench, it also includes establishing a data acquisition platform, inputting the defect information of the cable trench and the usage situation of the pipe orifices, and displaying the 3D model.

[0016] Compared with the prior art, the beneficial effects of the present invention are at least as follows: By setting surveying and mapping equipment in the on-site data acquisition link, the present invention conducts a full-range scan of the ground and underground of the cable trench, accurately obtains the longitude, latitude and position information of the cable trench through the high-precision measurement of lidar, combined with the assistance of spatial positioning and inertial measurement units; at the same time, the wide-angle camera can capture images of the cable trench, and the 3D point cloud data can present the 3D structure of the cable trench in detail; in the data transmission stage, a gateway device is set between the sensor device and the cloud platform for data encryption, protocol conversion, etc., which can reduce costs and enhance the security of data transmission.

[0017] In the data automatic modeling stage, make full use of data sources such as point clouds and panoramic photos, and construct a three-dimensional model of the cable shaft through advanced 3D point cloud reconstruction technology. These technologies can accurately restore the actual shape and structure of the cable shaft, providing a reliable model basis for subsequent data analysis and applications. Based on YOLOv8, establish a machine learning model. The YOLOv8 framework has powerful real-time object detection capabilities, can quickly and accurately analyze image data, identify the occupancy of profiles and pipe holes, as well as potential defects underground, providing important basis for the maintenance and management of cable shafts. In the data synchronization and backhaul link, with the powerful functions of the i-State Grid mobile application, ensure that the predicted underground detection data can be transmitted to the intranet desktop in a timely and accurate manner for subsequent data processing and analysis, improving work efficiency and data timeliness. Brief Description of the Drawings

[0018] Figure 1 It is a step flow chart of the mapping and recording method for underground cables and cable trenches based on artificial intelligence of the present invention; Figure 2 It is a comparison diagram before and after cropping of CloudCompare of the present invention; Figure 3 It is a data annotation diagram of the present invention. Detailed Embodiment

[0019] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0020] It can be understood that the terms "first", "second", etc. used in the present application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of the present application, the first xx script may be referred to as the second xx script, and similarly, the second xx script may be referred to as the first xx script.

[0021] As Figure 1 shown, a mapping and recording method for underground cables and cable trenches based on artificial intelligence includes: S1: Based on surveying equipment, scan the ground and underground of the cable shaft to obtain the first transmission information of the cable shaft, and the first transmission information includes three-dimensional point cloud data and image data.

[0022] Specifically, to improve the quality and efficiency of underground cable mapping work and reduce the work risks of personnel going down into wells, the present invention uses intelligent devices to replace personnel going down into wells. The mapping device consists of multiple sensor devices, including lidar devices and wide-angle cameras, and may also include spatial positioning devices such as GPS receivers, and inertial measurement devices such as accelerometers and gyroscopes, which are used to measure and record the movement and direction of the device, etc. Through the high-precision measurement of lidar, combined with the assistance of spatial positioning and inertial measurement units, the longitude, latitude and position information of the cable well can be accurately obtained; at the same time, the wide-angle camera can capture images of the cable well, and the three-dimensional point cloud data can present the three-dimensional structure of the cable well in detail, collecting multiple types of data at one time, providing a basis for subsequent artificial intelligence recognition and three-dimensional modeling work, and replacing field personnel for down-well mapping.

[0023] S2: Set the data transmission method, and transmit the first transmission information to the data platform based on the data transmission method and store it in the database.

[0024] Specifically, in order to transmit the sensor data obtained by the sensor device to the data cloud platform, certain configuration and programming work needs to be carried out on the device, which may include setting network connection parameters, authentication information of the cloud platform, data transmission protocols, etc. If each device needs to be configured separately, then as the number of devices increases, the total cost will also increase significantly. Therefore, the present invention sets a data transmission method and provides a common gateway device, which provides an intermediate layer between the terminal and the cloud platform, encrypts and converts the protocol of the data before it is transmitted to the cloud platform, enhancing the security of data transmission and reducing the cost.

[0025] S3: Extract the three-dimensional point cloud data from the first transmission information in the database, perform preliminary cropping on the three-dimensional point cloud data, denoise the cropped three-dimensional point cloud data based on the RANSAC algorithm to obtain standard point cloud data, obtain the spatial information of the cable well based on the standard point cloud data, and use 3D point cloud reconstruction technology to reconstruct the model of the cable well to obtain the three-dimensional model of the cable well.

[0026] Specifically, CloudCompare software is used to perform preliminary cropping on the three-dimensional point cloud data. CloudCompare is a software dedicated to the editing and processing of 3D point clouds (and triangular meshes), which can load a variety of open point cloud formats and crop the redundant point clouds other than the cable well. Figure 2It is a comparison diagram before and after cropping, showing the point cloud of the entire cable well. The more detailed the cropping, the more accurate the measurement of the cable well dimensions will be later. The RANSAC algorithm is used to identify the three-dimensional point cloud data after cropping, removing the noise points and outliers in the dataset, thereby obtaining a cleaner dataset, improving the recognition accuracy. And the RANSAC algorithm is used to extract the contour lines of each plane and segment them. This step helps to identify and separate different parts of the cable well, such as the well wall, the bottom of the well, and the well cover, etc. Calculate the dimensions of the bounding box of the standard point cloud data, and use the dimensions of the point cloud bounding box as the dimensions of the cable well, then the automatic measurement of the length, width, and height of the cable well can be realized. Obtain the dense point cloud underground, and finally perform mesh reconstruction on the dense point cloud to obtain the three-dimensional model of the cable well.

[0027] S4: Extract the image data from the first transmission information in the database, identify the key features in the image data, label the key features, establish a machine learning model based on the YOLOv8 algorithm. The machine learning model is used to detect the underground defects of the cable well and identify the occupancy of the pipe openings. Set the learning rate of the model, use the labeled key features as the training set to train the machine learning model, evaluate the trained machine learning model to obtain the accuracy of the model. If the accuracy is greater than the first threshold, use the predicted underground defect data and the identification data of whether the pipe opening is occupied as the underground detection data, and label the underground detection data on the three-dimensional model.

[0028] Specifically, in order to help the machine learning model understand the information in the picture, by labeling the key features in the picture, such as objects, scenes, attributes, as well as the occupancy of the cross-section and pipe openings, and the defect information underground of the cable well, it provides accurate data support for the training of the model. Through a labeling software, such as CPS software, label the occupancy of the pipe openings. As Figure 3 shown, it is a data labeling diagram. In the diagram, the green square represents unoccupied, and the red represents occupied.

[0029] YOLOv8 (YouOnlyLookOnce version 8) is a deep learning framework specifically designed for real-time object detection tasks. With its real-time performance, high accuracy, multi-scale prediction, adaptive anchor boxes, strong feature extractor, versatility, model selection for different scales, and excellent evaluation metrics, YOLOv8 has become a very powerful and flexible object detection framework with broad application prospects in multiple fields. Establishing a machine learning model based on the YOLOv8 algorithm can quickly and accurately analyze image data, identify the occupancy of profiles and pores, as well as potential defects underground, providing important basis for the maintenance and management of cable wells. Set the learning rate of the model, continuously adjust the learning rate during the model training process, use the training set to train the model, and continuously update the parameters of the model through the backpropagation algorithm to minimize the loss function and improve the performance and stability of the model. Evaluate the trained model, and deploy the machine learning model with an accuracy greater than the first threshold to the actual underground monitoring system for real-time defect detection and analysis.

[0030] As a preferred technical solution of the present invention, transmitting the first transmission information to the data platform and storing it in the database includes the following steps: Encapsulate the first transmission information into a data packet, add header information before the data packet. The header information includes the port number of the communication protocol used by the surveying and mapping device. Define the data packet with the added header information as the second transmission data. Set up a gateway device and transmit the second transmission data to the gateway device. The gateway device includes a data connection unit and a data queuing unit. The data connection unit removes the header information to extract the data packet, generates an encryption key, encrypts the data packet, assigns label information to the encrypted data packet, and defines it as the third transmission data.

[0031] The data queuing unit sets up a queue list, which has a corresponding relationship between multiple queues and label information. The data queuing unit stores the third transmission data in the corresponding queue based on the queue list, and converts the third transmission data into the fourth transmission data. The fourth transmission data is data that can be received by the data cloud platform after protocol conversion and format conversion. The data cloud platform obtains the fourth transmission data from the data queuing unit, decrypts the fourth transmission data using the key corresponding to the encrypted data, and stores the decrypted fourth transmission data in the database.

[0032] Specifically, for example, the first transmission information includes point cloud data information acquired by the laser radar, the point cloud data information is encapsulated into a data packet, header information is added before each data packet, the communication protocol used by the laser radar is UDP, wherein the port number of the MSOP package is 6699, and the port number of the DIFOP package is 7788. The specific port number is determined according to the specific model and configuration of the device, the header information (7788 or 6699) and the data packet are defined as the second transmission information, the second transmission information is transmitted to the gateway device, a data receiving unit is set in the gateway device, which is used to add label information to the data and perform encryption processing before the data is transmitted to the cloud platform. The label information is a unique identifier. The data processed by the data receiving unit is then transmitted to the data queuing unit, the data queuing unit sorts the data according to certain groups, so that the data is transmitted to the cloud platform in an orderly manner, and before the data is transmitted, the protocol and format of the data are converted into the protocol and data format received by the cloud interface. Due to the encryption and protocol conversion functions in the gateway device provided by the present invention, the IoT terminal does not need to have these functions, thereby reducing the communication volume and hardware cost of the IoT terminal.

[0033] Obtaining standard point cloud data includes the following steps: Step S31: Based on the RANSAC algorithm, a part of data is randomly selected from the three-dimensional point cloud data as the first sample data, and a geometric model of the bottom or side of the cable well is generated based on the first sample data. The vertical distance from each three-dimensional point cloud data in the first sample data to the geometric model is calculated. If the vertical distance is less than the second threshold, the three-dimensional point cloud data is defined as the first type of point cloud data, otherwise it is defined as the second type of point cloud data. The number of the first type of point cloud data and the second type of point cloud data in the first sample data are calculated respectively, and defined as the first number and the second number, and the second type of point cloud data in the first sample data is eliminated.

[0034] Step S32: Based on the RANSAC algorithm, randomly select a part of the data from the remaining three-dimensional point cloud data as the second sample data, repeat step S31, and calculate the ratio of the second number to the first number after each iteration. If the ratio is less than the third threshold, the denoising process of the RANSAC algorithm ends, and the remaining first-category point cloud data is defined as standard point cloud data.

[0035] Specifically, the RANSAC algorithm is used to randomly extract representative samples from the three-dimensional point cloud data that can be used for model fitting, that is, the first sample data. According to the first sample data, a geometric model of the bottom or side of the cable shaft is fitted. The vertical distance is used to evaluate the proximity of each three-dimensional point cloud data to the fitted model. If the vertical distance is less than the second threshold, the three-dimensional point cloud data is defined as the first type of point cloud data (i.e., inlier, considered to be part of the model), otherwise it is defined as the second type of point cloud data (i.e., outlier, considered to be noise or outliers). The second type of point cloud data in the first sample data is removed to remove those data points that are inconsistent with the model, thereby purifying the data set. Based on the RANSAC algorithm, a part of the remaining three-dimensional point cloud data is randomly selected as the second sample data, and step S31 is repeated. Calculate the ratio of the second quantity to the first quantity after each iteration. The second type of point cloud data will become less and less as the number of iterations increases. Therefore, when the ratio is less than the preset third threshold, for example, the third threshold is 0.1, indicating that the proportion of the second type of point cloud data in the sample data is getting smaller and the proportion of the first type of point cloud data is getting larger. Therefore, the remaining first type of point cloud data is defined as the standard data.

[0036] Obtaining the spatial information of the cable shaft includes the following steps: Obtain the maximum coordinate value and the minimum coordinate value of the standard point cloud data in the X, Y, and Z three coordinate axis directions, calculate the difference between the maximum coordinate value and the minimum coordinate value on each coordinate axis, and define them as the first length, the first width, and the first height respectively. Based on the first length, the first width, and the first height, construct a bounding box of the standard point cloud data, and obtain the spatial information of the cable shaft based on the size of the bounding box. The spatial information includes the length, width, and height of the cable shaft.

[0037] Specifically, by calculating the maximum and minimum coordinate values of an object in three-dimensional space along each coordinate axis, an AABB bounding box that can tightly enclose the object can be determined. The length, width, and height of this bounding box are the differences between the maximum and minimum values along each coordinate axis, so as to calculate the length, width, and height of the cable shaft, and various parameter information of the cable can be automatically obtained in this way.

[0038] Obtaining the three-dimensional model of the cable shaft includes the following steps: Perform Mesh reconstruction on the standard point cloud data, convert the standard point cloud data into a mesh model with multiple faces, and smooth the mesh model to generate the three-dimensional model of the cable shaft.

[0039] Specifically, triangulation is performed on the standard point cloud data, that is, the standard point cloud is converted into a mesh model composed of multiple triangular patches. This process can be achieved through various algorithms such as Delaunay triangulation. Through triangulation, a preliminary mesh model is generated. This model may contain a large number of triangular patches and requires further optimization and smoothing. Remove redundant patches, perform smoothing, and ensure the correct topological structure of the model. Verify the 3D model by comparing it with the actual measurement data to ensure that the size and shape of the model match the actual cable shaft.

[0040] Building a machine learning model based on the YOLOv8 algorithm includes the following steps: The machine learning model classifies the downhole defect information to obtain multiple defect categories. Define the defects predicted by the machine learning model as the first label and the actual downhole defects as the second label. Set the loss function of the YOLOv8 algorithm and calculate the difference value between the first label and the second label based on the loss function. The loss function is: , where is the prediction probability of the machine learning model for each defect category, is the coefficient used to balance the weights of positive and negative samples. Positive class samples refer to the situation where there are defects, and negative class samples refer to the situation where there are no defects. is the parameter for adjusting the sample weights, is the difference value. Update the model parameters of the machine learning model based on the difference value. When the difference value is less than the fourth threshold, the machine learning model is constructed.

[0041] Specifically, first, use the labeled dataset (including images of downhole defects and their corresponding labels) to train the YOLOv8 model. The model is trained by learning the features and patterns of these data with the aim of being able to identify and classify different defect categories. Evaluate the accuracy of the model prediction based on the difference value between the first label and the second label. For example, the smaller the difference value, the more accurate the prediction of the machine learning model. When the difference value is greater than the preset fourth threshold, update the model parameters of the machine learning model until the difference value is less than the fourth threshold. At this time, the model has achieved satisfactory performance and the machine learning model is constructed.

[0042] Setting the learning rate of the model includes the following steps: Continuously adjust the learning rate of the model training process based on the first formula , and the first formula is: , where is the initial learning rate, is the preset decay rate, is the number of training times, and the machine learning model is trained based on the learning rate.

[0043] Specifically, the learning rate of the model can be effectively controlled to gradually decrease during the training process, thereby improving the convergence speed and final performance of the machine learning model.

[0044] Evaluating the trained machine learning model to obtain the accuracy of the model includes the following steps: Use the test set to evaluate the machine learning model and calculate the accuracy based on the second formula , and the second formula is: , where is the number of samples correctly predicted as positive by the model, is the number of samples correctly predicted as negative by the model, is the number of samples incorrectly predicted as positive by the model, is the number of samples incorrectly predicted as negative by the model.

[0045] After detecting the defect information of the cable shaft, it also includes establishing a data collection platform, inputting the defect information of the cable shaft and the usage of the pipe openings, and displaying the 3D model.

[0046] Specifically, by leveraging the powerful functions of the i-State Grid mobile application, the real-time transmission of surveying and mapping data can be achieved, ensuring that the predicted downhole detection data can be transmitted to the data cloud platform on the intranet desktop in a timely and accurate manner, and displaying the generated 3D model to improve work efficiency and data timeliness.

[0047] It should be understood that although the steps in the flowcharts of the embodiments of the present invention are shown in sequence according to the arrows, these steps do not necessarily have to be executed in the order indicated by the arrows. Unless otherwise clearly stated in this article, the execution of these steps does not have a strict order limit, and these steps can be executed in other orders. Moreover, at least some of the steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages do not necessarily have to be executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages does not necessarily have to be sequential, but can be executed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

[0048] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The above program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0049] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0050] The above embodiments only represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can be made, and these all belong to the protection scope of the present invention.

[0051] The above is only the preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for mapping and recording underground cables and trenches based on artificial intelligence, characterized in that: The method comprises the following steps: S1: Scanning the surface and underground of the cable well based on surveying and mapping equipment to obtain first transmission information of the cable well, where the first transmission information includes three-dimensional point cloud data and image data; S2: Setting a data transmission mode, transmitting the first transmission information to a data platform based on the data transmission mode, and storing the information in a database; S3: extracting three-dimensional point cloud data from the first transmission information in the database, preliminarily cropping the three-dimensional point cloud data, denoising the cropped three-dimensional point cloud data based on the RANSAC algorithm to obtain standard point cloud data, obtaining spatial information of the cable well based on the standard point cloud data, and reconstructing a model of the cable well using 3D point cloud reconstruction technology to obtain a three-dimensional model of the cable well; S4: extracting image data from the first transmission information in the database, identifying key features in the image data, annotating the key features, establishing a machine learning model based on the YOLOv8 algorithm, the machine learning model being used to detect downhole defects in the cable well and identify the occupancy of the pipeline opening, setting the learning rate of the model, training the machine learning model using the annotated key features as a training set, evaluating the trained machine learning model to obtain the accuracy of the model, and if the accuracy is greater than a first threshold, using the predicted downhole defect data and the identification data of whether the pipeline opening is occupied as downhole detection data, and annotating the downhole detection data on the three-dimensional model.

2. The method according to claim 1, characterized in that Transmitting the first transmission information to the data platform and storing it in the database includes the following steps: Encapsulating the first transmission information into a data packet, adding header information before the data packet, the header information including the port number of the communication protocol used by the surveying and mapping equipment, defining the data packet with the header information added as second transmission data, setting a gateway device, transmitting the second transmission data to the gateway device, the gateway device including a data connection unit and a data queuing unit, the data connection unit removing the header information to extract the data packet, generating an encryption key, encrypting the data packet, assigning label information to the encrypted data packet, and defining it as third transmission data; The data queuing unit sets a queue table, and there is a correspondence between multiple queues and tag information in the queue table. The data queuing unit stores the third transmission data in the corresponding queue based on the queue table, and converts the third transmission data into fourth transmission data. The fourth transmission data is data that can be received by the data cloud platform after protocol conversion and format conversion. The data cloud platform obtains the fourth transmission data from the data queuing unit, decrypts the fourth transmission data using the key corresponding to the encrypted data, and stores the decrypted fourth transmission data in the database.

3. The method according to claim 1, characterized in that Obtaining standard point cloud data includes the following steps: Step S31: randomly selecting a portion of data from the three-dimensional point cloud data as first sample data based on the RANSAC algorithm, fitting and generating a geometric model of the bottom or side of the cable well based on the first sample data, calculating a vertical distance from each three-dimensional point cloud data in the first sample data to the geometric model, and if the vertical distance is less than a second threshold, defining the three-dimensional point cloud data as first-category point cloud data, otherwise defining it as second-category point cloud data, respectively calculating the number of first-category point cloud data and second-category point cloud data in the first sample data, defining them as a first number and a second number, and eliminating the second-category point cloud data in the first sample data; Step S32: Based on the RANSAC algorithm, randomly select a part of the data from the remaining three-dimensional point cloud data as the second sample data, repeat step S31, and calculate the ratio of the second number to the first number after each iteration. If the ratio is less than the third threshold, the denoising process of the RANSAC algorithm ends, and the remaining first-category point cloud data is defined as standard point cloud data.

4. The method according to claim 3, characterized in that Obtaining the spatial information of the cable well comprises the following steps: Obtain the maximum coordinate value and the minimum coordinate value of the standard point cloud data along the three coordinate axes of X, Y and Z, calculate the difference between the maximum coordinate value and the minimum coordinate value on each coordinate axis, define them as the first length, the first width and the first height respectively, construct a bounding box of the standard point cloud data based on the first length, the first width and the first height, and obtain the spatial information of the cable well based on the size of the bounding box, wherein the spatial information includes the length, width and height of the cable well.

5. The method according to claim 4, characterized in that Obtaining the three-dimensional model of the cable well comprises the following steps: The standard point cloud data is mesh reconstructed, the standard point cloud data is converted into a mesh model of multiple faces, the mesh model is smoothed, and a three-dimensional model of the cable well is generated.

6. The method according to claim 1, characterized in that Building a machine learning model based on the YOLOv8 algorithm includes the following steps: The machine learning model classifies the downhole defect information to obtain multiple defect categories, defines the defects predicted by the machine learning model as the first label, and defines the actual downhole defects as the second label. The loss function of the YOLOv8 algorithm is set, and the difference between the first label and the second label is calculated based on the loss function. The loss function is: ,in, is the predicted probability of the machine learning model for each of the defect categories, It is a coefficient used to balance the weights of positive and negative samples. Positive samples refer to cases where defects exist, and negative samples refer to cases where there are no defects. is a parameter to adjust the sample weight, The difference value is used to update the model parameters of the machine learning model based on the difference value. When the difference value is less than a fourth threshold, the machine learning model is constructed.

7. The method according to claim 1, characterized in that The key features include the occupancy of the profile and the pipe opening, and the defect information underground in the cable well.

8. The method according to claim 1, characterized in that Setting the learning rate of the model involves the following steps: Based on the first formula, the learning rate of the model training process is continuously adjusted , the first formula is: ,in, is the initial learning rate, is the preset decay rate, is the number of training times, and the machine learning model is trained based on the learning rate.

9. The method according to claim 6, characterized in that Evaluating the trained machine learning model to obtain the model's accuracy includes the following steps: The machine learning model is evaluated using a test set, and the accuracy is calculated based on a second formula , the second formula is: ,in, is the number of samples correctly predicted as positive by the model, The number of samples that the model correctly predicts as negative, is the number of samples that the model incorrectly predicts as positive. The number of samples that the model incorrectly predicts as negative.

10. The method according to claim 1, characterized in that After detecting the defect information of the cable well, the method also includes establishing a data collection and recording platform, recording the defect information of the cable well and the usage of the pipeline opening, and displaying the three-dimensional model.

Images