Cable trench construction auditing method and equipment based on robot image

Abstract

The invention discloses a cable trench construction auditing method and equipment based on robot images, and aims to solve the problems of large potential safety hazards, low efficiency, subjective and inaccurate data, difficulty in information tracing and single function of existing auxiliary equipment in traditional cable trench construction auditing. The equipment comprises a mobile robot platform, and a comprehensive sensing module, a calculation control module and a communication module which are mounted on the platform, the comprehensive sensing module comprises a binocular depth camera, a laser radar, a high-precision odometer, a multispectral illumination system, an inertial measurement unit and an optional gas sensor, and each module is electrically connected with the calculation control module. The method comprises the steps of task planning and robot deployment, robot autonomous navigation and collection of data with space coordinates, data preprocessing and fusion to generate a three-dimensional point cloud model, AI construction element recognition, automatic measurement and standard comparison, defect diagnosis and report generation, and data archiving and visual display. According to the invention, manual work is replaced by the robot, multi-sensor fusion, SLAM and AI analysis are combined, auditing security, high efficiency and precision are realized, and a traceable digital auditing file is formed.

Description

Technical Field

[0001] This invention relates to the field of power engineering construction quality inspection technology, specifically providing a method and equipment for cable trench construction auditing based on robot imaging. Background Technology

[0002] With the upgrading of the municipal power grid, the construction and maintenance of underground cable trenches has surged. The quality of their construction directly affects the safe and stable operation of the power grid, making auditing and acceptance work of paramount importance. Traditional cable trench construction audits mainly rely on manual methods. Auditors must enter narrow, dark trenches that may contain harmful gases and record the actual condition of key items such as civil structures and support installations through visual observation, manual measurement, and photography. From a safety perspective, cable trenches may pose risks such as oxygen deficiency, toxic and harmful gases, and structural collapse, seriously threatening the lives of auditors. From an efficiency perspective, manual measurement and recording are slow, requiring a significant investment of manpower and time for a cable trench several kilometers long, severely impacting the overall project progress. From a data accuracy perspective, visual observation and manual measurement are prone to errors, making it difficult to obtain high-precision quantitative data. Audit results are also influenced by personnel experience and work status, exhibiting strong subjectivity and insufficient objectivity. From an information management perspective, manual photography may suffer from omissions, poor angles, and insufficient clarity. Furthermore, the massive amount of image data lacks effective management methods and spatial location information correlation, making subsequent tracing, verification, and archiving extremely difficult. From an equipment functionality perspective, fixed monitoring limits the audit scope due to its immobility, while handheld pole-mounted devices cannot achieve autonomous coverage of the entire trench and lack precise positioning and measurement capabilities, still requiring substantial manual intervention. Summary of the Invention

[0003] To address the technical problem that existing technologies for cable trench construction auditing cannot form an automated, digital, and quantifiable unmanned audit loop, leading to personnel safety risks during audits, this invention provides a method and equipment for cable trench construction auditing based on robotic imaging.

[0004] This invention provides a method for auditing cable trench construction based on robot imagery, comprising the following steps: The robot's audit path is pre-planned, and the robot moves autonomously along the cable trench; During the robot's movement, the robot uses data acquired by LiDAR and inertial measurement unit to construct map data of the cable trench and estimate the robot's pose data using SLAM algorithm; the robot collects image data and depth data through binocular cameras; and the robot uses a multispectral lighting system for ambient lighting; wherein the map data, pose data, image data, and depth data are bound to spatial position coordinates; Map data, pose data, image data, and depth data are aligned and fused in a timestamp and spatial coordinate system to generate a 3D point cloud model of the cable trench. The 3D point cloud model and image data of the cable trench are input into a pre-trained deep learning model. The deep learning model identifies and segments the construction elements in the cable trench, and generates a list of elements with unique IDs and location information in the spatial coordinate system for each construction element. Automated measurement is performed on each construction element in the element list to obtain measurement results, and the measurement results are compared with a pre-set construction specification standard database to diagnose defects that do not conform to the construction specification standards. Defect points are labeled in the 3D point cloud model of the cable trench, and a structured audit report is generated.

[0005] Furthermore, the planning of the robot's audit path specifically involves planning the robot's approximate audit path in the background software based on the cable trench design drawings; after the system is started, it also includes the step of completing the self-test of each sensor of the audit robot.

[0006] Furthermore, it also includes the use of a lidar and an inertial measurement unit to construct a three-dimensional map of the cable trench through a synchronous positioning and mapping algorithm, and to estimate the robot pose through the synchronous positioning and mapping algorithm; and a binocular depth camera to acquire image data and depth data at specific intervals or continuously.

[0007] Furthermore, the system also includes the following: the map data generated by the SLAM algorithm is based on a global coordinate system; the pose data is the robot's real-time position and orientation relative to the global coordinate system; the image data and depth data acquired by the binocular camera are converted to the robot's body coordinate system through a pre-calibrated extrinsic parameter matrix, and uniformly registered to the global coordinate system according to the real-time pose data, so that the image data and its corresponding depth point cloud both have global coordinates.

[0008] Furthermore, the pre-trained deep learning model includes construction elements within the cable trench identified and segmented by a convolutional neural network, including cable supports, grounding flat irons, high-voltage cables, and fireproof partitions.

[0009] Furthermore, the automated measurement of construction elements specifically includes: for cable supports, measuring the spacing between adjacent cable supports, the verticality of the cable support installation, and the flatness of the top surface of the cable support; for grounding flat irons, measuring the welding overlap length and bending radius of the grounding flat irons; and for high-voltage cables, detecting whether there is damage to the surface of the high-voltage cable and estimating the damaged area.

[0010] Furthermore, the annotation of defect points in the 3D point cloud model specifically involves highlighting defect points in the 3D point cloud model; the structured audit report includes an overview of the cable trench, an audit data overview, a defect list, and a link to browse the 3D point cloud model. The overview information includes the trench location and audit time, the audit data overview includes the total length of the cable trench and the number of supports, and the defect list includes the defect type, defect location, defect images, and a comparison of measured values ​​with standard values.

[0011] Furthermore, it also includes a cloud platform for storage; and enables the visualization of 3D models and audit results, specifically allowing users to remotely view the 3D point cloud model of the cable trench and the corresponding audit results through a web browser.

[0012] Furthermore, the robot is equipped with a gas sensor. During the robot's autonomous movement along the cable trench, the gas sensor is also used to detect the oxygen concentration, combustible gas concentration, and toxic gas concentration in the cable trench. The data detected by the gas sensor is also bound to the spatial position coordinates and transmitted to the computing and control module.

[0013] The present invention also provides a cable trench construction auditing device based on robot imaging, applicable to any of the described cable trench construction auditing methods based on robot imaging, including a mobile robot platform, an integrated sensing module, a computing control module, and a communication module; The integrated sensing module, the computing control module, and the communication module are all installed on the mobile robot platform, and the integrated sensing module and the communication module are electrically connected to the computing control module respectively. The integrated sensing module includes a binocular depth camera, lidar, high-precision odometer, multispectral illumination system, inertial measurement unit, and optional gas sensor. The binocular depth camera is located at the front end of the mobile robot platform, the lidar is located at the top of the mobile robot platform, the high-precision odometer is integrated into the mobile chassis of the mobile robot platform, the multispectral illumination system is arranged around the binocular depth camera, the inertial measurement unit is mounted on the mobile robot platform, and the optional gas sensor is mounted on the mobile robot platform. The binocular depth camera, lidar, high-precision odometer, multispectral illumination system, inertial measurement unit, and optional gas sensor are electrically connected to the computing and control module.

[0014] Beneficial effects This invention provides a method and equipment for cable trench construction auditing based on robot imaging. By advancing the technology of cable trench construction auditing, it improves the safety of auditing work, effectively avoids the risks of harmful gases and structural collapse faced by traditional manual trenching, protects the health and safety of auditors, eliminates safety hazards from the working environment, and allows auditing work to be carried out in a safer environment. Simultaneously, it significantly improves auditing efficiency, eliminating the cumbersome process of manual measurement and recording, reducing manpower input, shortening the audit cycle, and preventing the overall project progress from being affected by excessively long audit times, thus enabling auditing work to more efficiently serve the advancement of cable trench construction projects. Regarding data accuracy, advanced technology can acquire high-precision quantitative data, reducing errors caused by visual observation and manual measurement, minimizing the impact of personnel experience and condition on audit results, improving the objectivity and reliability of audit data, and providing stronger data support for audit conclusions. Furthermore, it optimizes information management, avoiding problems such as omissions, poor angles, and insufficient clarity that may occur with manual photography, achieving standardized management of audit image data and accurate correlation of spatial location information, facilitating later traceability, review, and archiving, and improving the standardization and convenience of information management. Furthermore, by integrating technologies, we can address the issue of limited functionality in existing auxiliary equipment, expand the scope of audits, reduce manual intervention, and promote the transformation of audit work towards automation, digitalization, and intelligence. This will comprehensively improve the overall quality and level of cable trench construction audits, provide stronger guarantees for the safe and stable operation of urban power grids, facilitate the high-quality advancement of urban power grid construction and maintenance, and meet the needs of long-term stable development of urban power grids. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating the working steps of the cable trench construction audit method based on robot imaging provided in this embodiment of the invention; Figure 2 A schematic diagram of a cable trench construction auditing device based on robot imaging provided in another embodiment of the present invention; Figure 3 A physical image of a cable trench construction auditing device based on robot imaging, provided for another embodiment of the present invention.

[0017] Reference numerals: 1. Mobile robot platform; 2. Integrated sensing module; 3. Computation and control module; 4. Communication module. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] It should be noted that if the embodiments of the present invention involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a specific posture. If the specific posture changes, the directional indicators will also change accordingly.

[0020] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Thus, features defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the use of "and / or" or "and / or" throughout the text implies three parallel solutions. For example, "A and / or B" includes solution A, solution B, or a solution where both A and B are satisfied. Furthermore, the technical solutions of the various embodiments can be combined, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0021] In existing technologies, auditing the construction of underground cable trenches mainly relies on manual methods. Auditors must enter narrow, dark trenches that may contain harmful gases to visually inspect, manually measure, and photograph key items. This method poses safety risks, such as oxygen deficiency, toxic gases, and structural collapse threatening personnel safety; it is inefficient, requiring a large amount of manpower and time for cable trenches several kilometers long; data accuracy is insufficient, as manual measurements are prone to errors and are influenced by subjective factors; information management is difficult, as image data lacks spatial correlation, making tracing and archiving challenging; and equipment functionality is limited, with fixed monitoring unable to cover the entire trench, and handheld devices still requiring manual intervention and lacking precise positioning capabilities.

[0022] To address the aforementioned issues, this paper considers using robots to replace human intervention in hazardous environments to mitigate the security risks of traditional auditing methods; to address inefficiencies and data errors, it explores automated data acquisition through multi-sensor fusion; to address disorganized information management, it investigates methods for binding data to spatial coordinates; and to address insufficient equipment functionality, it designs a mobile platform with autonomous navigation and precise positioning. By integrating LiDAR, visual sensors, and AI algorithms, a complete technical solution covering data acquisition, processing, analysis, and visualization is constructed.

[0023] refer to Figure 1 This embodiment provides a method for auditing cable trench construction based on robot imaging, including the following steps: S1. The robot's audit path is pre-planned, and the robot moves autonomously along the cable trench; S2. During the robot's movement, the robot uses data acquired by LiDAR and inertial measurement unit, and constructs map data of the cable trench and estimates the robot's pose data using SLAM algorithm; the robot collects image data and depth data through binocular cameras; the robot uses a multispectral lighting system for environmental illumination; wherein, the map data, pose data, image data, and depth data are bound to spatial position coordinates; S3. Align and fuse map data, pose data, image data, and depth data in the timestamp and spatial coordinate system to generate a 3D point cloud model of the cable trench. S4. Input the 3D point cloud model and image data of the cable trench into the pre-trained deep learning model. The deep learning model identifies and segments the construction elements in the cable trench and generates a list of elements with unique IDs and location information in the spatial coordinate system for each construction element. S5. Perform automated measurement for each construction element in the element list, obtain the measurement results, and compare the measurement results with a pre-set construction specification standard database to diagnose defects that do not conform to the construction specification standards. S6. Mark defect points in the 3D point cloud model of the cable trench and generate a structured audit report. It should be noted that: Audit path planning refers to determining the robot's route based on the cable trench design drawings, which can be implemented using path planning algorithms to ensure coverage of the audit area. Autonomous navigation refers to building an environmental map in real time using LiDAR and inertial measurement units, specifically implemented using SLAM algorithms to provide a spatial reference for data acquisition. A binocular depth camera refers to an imaging device with stereo vision, specifically implemented using an industrial-grade binocular camera with adjustable baseline distance, used to acquire 3D spatial information. A multispectral lighting system refers to a supplementary lighting device containing multiple wavelength light sources, specifically implemented using an LED array, automatically adjusting the supplementary lighting parameters according to the ambient illuminance. A 3D point cloud model refers to a spatial model that integrates multi-source data, specifically generated using point cloud registration algorithms, preserving color and geometric information. A deep learning model refers to a trained neural network, specifically implemented using a convolutional neural network architecture, used to identify construction elements such as cable supports. Standard compliance analysis refers to comparing measurement data with standard thresholds, specifically implemented using a rule engine to automatically determine whether the construction quality meets standards.

[0024] Specifically, during the robot deployment phase, a pre-defined path ensures full audit coverage, and sensor self-checks guarantee system reliability. During movement, LiDAR scans the environment to generate point clouds, and the inertial measurement unit records the motion trajectory. Both work together to construct a map and locate the robot in real time using a SLAM algorithm. A binocular camera acquires high-definition images and depth maps at a fixed frequency, and multispectral illumination automatically supplements light based on ambient brightness to eliminate shadows. All acquired data is bound to spatial coordinates, forming a spatiotemporally aligned data stream. In the preprocessing phase, timestamp matching and coordinate system transformation fuse point clouds, images, and pose data into a 3D model with color information. The AI ​​model performs joint analysis of the 3D model and images, accurately segmenting elements such as cable supports and recording their spatial positions. The automated measurement module calls appropriate algorithms for different element types; for example, the point cloud spacing measurement algorithm is used to calculate support spacing, and an image segmentation algorithm is used for cable damage detection. Measurement results are automatically compared with a standard database, generating a defect diagnosis report and annotating it in the 3D model. Finally, all data is uploaded to a cloud platform, supporting web-based 3D visualization queries.

[0025] In some embodiments, planning the robot's audit path specifically involves planning the robot's approximate audit path in the background software based on the cable trench design drawings; after the system is started, it also includes the step of completing the self-test of each sensor of the audit robot.

[0026] It should be noted that cable trench design drawings refer to design documents prepared before construction, including the cable trench's direction, dimensions, and key node locations. These can be implemented using CAD or BIM format electronic drawings and serve as a benchmark for robot path planning. The backend software refers to a computer program with path planning capabilities, typically implemented using a GIS-based path planning module. It generates the robot's path by parsing spatial data from the design drawings. The self-checking of the audit robot's sensors involves initializing the status of devices such as the LiDAR, binocular depth camera, and inertial measurement unit. This is achieved by sending test commands and receiving feedback signals, ensuring the sensors are in normal working order before the audit begins. During the task planning phase, the backend software imports the cable trench design drawings, automatically extracts the trench direction, turning radius, and manhole location information, and generates an audit path including the start point, end point, and key coordinates. This path serves as the robot's baseline for navigation, avoiding path deviations caused by manual experience-based planning. Upon robot startup, it sequentially performs self-checks on the LiDAR's point cloud generation function, the binocular camera's imaging clarity, and the inertial measurement unit's acceleration detection accuracy. If any sensor abnormalities are detected, an alarm is triggered, and the task is terminated. The system is restarted after the fault is resolved.

[0027] In some embodiments, the lidar and inertial measurement unit construct a two-dimensional or three-dimensional map of the cable trench through a synchronous positioning and mapping algorithm, and estimate the robot pose through the synchronous positioning and mapping algorithm; the binocular depth camera acquires high-definition color images and depth information at specific intervals or continuously.

[0028] Simultaneous localization and mapping (SLAM) algorithms refer to algorithms that generate environmental maps in real time and calculate the device's own position by fusing data from multiple sensors. Specifically, they can be implemented using graph-optimized SLAM frameworks or filtering algorithms to solve localization and mapping problems in cable trenches where there is no GPS signal. LiDAR (Light Detection and Ranging) is a sensor that acquires object distance information by emitting laser beams and receiving reflected signals. Specifically, it can be implemented using rotating or solid-state LiDAR to generate point cloud data of the cable trench structure. An inertial measurement unit (IMU) is a sensor combination consisting of accelerometers and gyroscopes. Specifically, it can be implemented using MEMS or fiber optic gyroscope technology to provide angular velocity and acceleration data during robot movement. A binocular depth camera is a visual sensor that calculates object distance through the parallax of two cameras. Specifically, it can be implemented using a global shutter camera and stereo matching algorithms to simultaneously acquire environmental images and depth information. Specific interval or continuous acquisition modes refer to acquisition modes that are dynamically adjusted according to storage capacity or processing power. Specifically, they can be implemented using fixed time interval triggering or real-time streaming to balance data integrity and system resource consumption. During robot movement, the LiDAR continuously scans the surrounding environment to generate point cloud data, while the inertial measurement unit (IMU) collects motion parameters in real time. These two types of data are input into a simultaneous localization and mapping (SLAM) algorithm for joint optimization, generating a 2D or 3D map of the cable trench and continuously correcting the robot's pose. The binocular depth camera collects data according to a preset strategy, such as triggering a shot at fixed intervals or maintaining continuous shooting mode. The acquired high-definition images and depth information are correlated with timestamps and spatial coordinates to form structured data with 3D spatial information. Data fusion between the LiDAR and IMU is achieved through Kalman filtering to smooth the motion trajectory, while the depth information from the binocular camera is generated using a stereo matching algorithm and aligned with the LiDAR point cloud in a unified coordinate system.

[0029] In some embodiments, laser point cloud data, visual image data, and pose data are aligned and fused in a timestamp and spatial coordinate system to generate a high-precision 3D point cloud model of a cable trench with color information and in a unified coordinate system.

[0030] Laser point cloud data refers to a dense set of 3D coordinate points obtained by scanning the surface of cable trenches with a lidar system. This can be achieved using a rotating lidar system in a horizontal scanning manner, used to construct the 3D spatial contour of the cable trench. Visual image data refers to a sequence of high-resolution color images acquired by a binocular depth camera. This can be achieved using a global shutter sensor capturing images at a rate of 30 frames per second, used to record the surface texture and color features of the cable trench. Timestamp alignment refers to matching data collected by different sensors according to a unified time reference. This can be achieved by using hardware synchronization signals to trigger synchronous acquisition by each sensor, ensuring consistency in the time dimension across different data sources. Spatial coordinate system fusion refers to converting laser point cloud, visual image, and robot pose data to the same spatial coordinate system. This can be achieved using pose data output from the robot's odometry and inertial measurement unit as the conversion reference, eliminating spatial deviations between multi-source data. Color information refers to mapping the RGB values ​​of the visual image to the corresponding coordinates in the 3D point cloud model. This can be achieved by matching pixel colors with 3D coordinates using point cloud coloring algorithms, enhancing the visualization effect of the 3D model.

[0031] LiDAR and a binocular depth camera respectively acquire geometric structure data and surface texture data of the cable trench, while an inertial measurement unit and odometry continuously record the robot's movement trajectory. In the data preprocessing stage, all sensor data are marked with precise timestamps, and interpolation algorithms are used to eliminate offsets caused by time delays. Subsequently, the LiDAR point cloud data is transformed into a global coordinate system using robot pose information, and the visual image data is projected into the same coordinate system using camera calibration parameters. Finally, the geometric point cloud and texture color are matched pixel-level to form a 3D point cloud model containing spatial coordinates, geometric shape, and surface color attributes. This model can fully reflect the actual state of the cable trench, providing an accurate data foundation for subsequent construction element identification.

[0032] In some embodiments, the pre-trained deep learning model includes construction elements within the cable trench identified and segmented by a convolutional neural network, including cable supports, grounding flat irons, high-voltage cables, and fireproof partitions.

[0033] A pre-trained deep learning model refers to a neural network model trained on a large number of labeled cable trench construction element samples. Specifically, it can be implemented using a convolutional neural network based on the U-Net architecture. This network achieves image feature extraction and pixel-level segmentation through multi-layer convolution and deconvolution operations. Cable supports, grounding flat irons, high-voltage cables, and fireproof partitions are key structural components that must be installed during cable trench construction. Specifically, these components can be located by combining geometric feature analysis of a 3D point cloud model with image texture recognition. For example, cable supports can be located by identifying the geometric features of vertical columns and beams. After a 3D point cloud model and high-resolution color imagery are input into a convolutional neural network, the network first extracts the texture features of the imagery and the geometric features of the point cloud through convolutional layers. Then, a feature fusion module correlates the data from the two modalities. In the segmentation stage, the network outputs the probability of the construction element category corresponding to each pixel. Finally, a post-processing algorithm generates a list of elements with unique IDs and 3D coordinates. For example, fireproof partitions can be identified by detecting their rectangular outline features and the color features of the fireproof coating, while high-voltage cables are segmented by identifying cylindrical geometric features and the texture features of the insulation layer.

[0034] In some embodiments, performing automated measurements on construction elements specifically includes: for cable supports, measuring the spacing between adjacent cable supports, the verticality of the cable support installation, and the flatness of the top surface of the cable support; for grounding flat irons, measuring the welding overlap length and bending radius of the grounding flat irons; for high-voltage cables, detecting whether there is damage to the sheath of the high-voltage cable and estimating the damaged area; the pre-set construction specification standard database is a national or industry construction specification standard database.

[0035] The spacing between adjacent cable supports refers to the straight-line distance between two adjacent supports. This can be measured non-contactly using lidar or vision sensors to ensure the support distribution meets safety spacing standards. The verticality of the cable support installation refers to the angular deviation of the support from the horizontal plane. This can be achieved using inertial measurement units or 3D point cloud model analysis to determine if the support installation is tilted. The flatness of the cable support top surface refers to the degree of undulation on the support's top surface. This can be calculated using a 3D point cloud model surface fitting algorithm to assess the uniformity of the support's bearing surface. The weld overlap length of the grounding flat iron refers to the length of the overlapping welded portion of two flat iron sections. This can be measured using image recognition technology combined with depth information to verify whether the welding quality meets conductivity requirements. The bending radius of the grounding flat iron refers to the inner arc radius of the bent portion of the flat iron. This can be calculated by extracting the geometric features of the bent section from a 3D point cloud model to determine whether the bending process meets mechanical strength requirements. High-voltage cable sheath damage detection refers to identifying defects such as tears and cracks in the cable's outer sheath. This can be achieved using deep learning models for semantic segmentation of high-definition images to assess insulation performance risks. A pre-built construction specification standard database refers to a structured dataset that stores the threshold values ​​of construction parameters stipulated by the state or industry. It can be implemented using a relational database or a cloud-based data table to provide benchmark values ​​for automated comparison.

[0036] After identifying the construction elements, for cable support elements, the spacing between adjacent supports was calculated using the spatial coordinates of the 3D point cloud model. The vertical axis was fitted using the point cloud distribution of the support columns, and the angle between the axis and the direction of gravity was calculated to obtain the verticality. The surface standard deviation was calculated after planar fitting of the point cloud on the top surface of the support to assess flatness. For grounding flat iron elements, the welding area was located through image recognition, and the overlap length was calculated using depth information. Curve fitting was performed on the 3D point cloud of the bent parts to calculate the minimum radius of curvature. For high-voltage cable elements, a semantic segmentation model was used to identify damaged areas on the outer sheath, and the actual projected area of ​​the damaged area was calculated using depth information.

[0037] In some embodiments, step S6, marking defect points in the 3D point cloud model specifically means marking defect points in the 3D point cloud model in a highlighted manner; the structured audit report includes an overview of the cable trench, an audit data overview, a defect list, and a link to browse the 3D point cloud model, wherein the overview information includes the trench location and audit time, the audit data overview includes the total length of the cable trench and the number of supports, and the defect list includes defect type, defect location, defect image, and comparison of measured values ​​with standard values.

[0038] Highlighting defect points involves highlighting abnormal areas in a 3D model using color differences or flashing effects. This can be achieved by overlaying semi-transparent color blocks with polygonal outlines, making defect locations visually easily identifiable. A structured audit report organizes audit results into a standardized, machine-readable document according to a pre-defined template. This can be stored in JSON or XML format and linked to a database for easy subsequent querying and statistical analysis. A 3D point cloud model browsing link generates an online URL after uploading the model file to the cloud. This can be achieved using WebGL technology for browser-side 3D rendering, supporting rotation, scaling, and layered viewing. Specifically, when the system detects non-compliance with specifications in construction elements, it first automatically matches a preset highlighting style based on the defect type. For example, if the welding length of the grounding flat iron is insufficient, a red semi-transparent mark is overlaid at the corresponding location. Subsequently, the system extracts the spatial coordinates, associated images, and measurement data of the defect, integrating them with basic information such as the trench location and audit time into a structured report. Users can directly jump to the 3D model interface by clicking links in the report to verify the defect by comparing the defect images with the measurement data.

[0039] Preferably, in some embodiments, the data is also transmitted to a storage platform, which is a cloud platform; the visualization of the 3D model and audit results is specifically achieved by the user remotely viewing the 3D point cloud model of the cable trench and the corresponding audit results through a web browser.

[0040] Remote viewing via web browser refers to accessing a visualization interface provided by a cloud platform through the HTTP protocol without requiring the user to install dedicated software. Specifically, it can utilize browsers like Chrome, Firefox, or Edge to load a 3D rendering engine based on WebGL or Three.js, enabling cross-platform and cross-device model interaction and data retrieval. This solves the problem of traditional audit result reliance on on-site equipment or dedicated software. During the data archiving phase, raw data, 3D point cloud models, and structured audit reports are uploaded to the cloud platform via a communication module. The cloud platform categorizes and stores the data and establishes indexes. In the visualization phase, the web service deployed on the cloud platform converts the 3D point cloud model into a lightweight format and provides model loading, viewpoint switching, and defect annotation viewing functions to the browser via API interfaces. Users log in to the cloud platform webpage by entering their authorized account and password, select the target cable trench audit project, and can directly rotate and zoom the 3D model in the browser, click on defect markers to view corresponding measurement data and specification comparison information, without needing to download a local client or go to the construction site.

[0041] refer to Figure 2 The present invention also provides another embodiment, a cable trench construction audit device based on robot imagery, including a mobile robot platform 1, a comprehensive sensing module 2, a computing and control module 3, and a communication module 4; the comprehensive sensing module 2, the computing and control module 3, and the communication module 4 are all installed on the mobile robot platform, and the comprehensive sensing module 2 and the communication module 3 are electrically connected to the computing and control module respectively; the comprehensive sensing module 2 includes a binocular depth camera, a lidar, a high-precision odometer, a multispectral illumination system, an inertial measurement unit, and an optional gas sensor; the binocular depth camera is located at the front end of the mobile robot platform, the lidar is located at the top of the mobile robot platform, the high-precision odometer is integrated into the mobile chassis of the mobile robot platform, the multispectral illumination system is arranged around the binocular depth camera, the inertial measurement unit is installed on the mobile robot platform 2, and the optional gas sensor is installed on the mobile robot platform 2; the binocular depth camera, lidar, high-precision odometer, multispectral illumination system, inertial measurement unit, and optional gas sensor are electrically connected to the computing and control module 3 respectively. The mobile robot platform 1 refers to a carrier device with autonomous mobility, specifically employing a tracked or wheeled chassis structure to adapt to the complex terrain within the cable trench. Its function is to provide stable support and mobility for sensors and computing units. The binocular depth camera is a device that acquires stereoscopic vision information through dual lenses, specifically using a synchronously triggered dual-camera module to collect high-definition images and depth data within the cable trench. Combined with a multispectral illumination system, it can maintain image clarity in low-light environments. The lidar is a sensor that measures distance by emitting laser beams, specifically employing a rotating or solid-state scanning device, used to construct a 3D point cloud model of the cable trench and assist in positioning. The high-precision odometer is a device that measures distance traveled through an encoder or visual odometer, specifically employing a solution integrating an optoelectronic encoder and an inertial measurement unit (IMU) for real-time recording of the robot's trajectory. The multispectral illumination system is a supplementary lighting device covering different wavelengths, specifically using an adjustable brightness LED array to provide uniform illumination when light is insufficient within the cable trench. The inertial measurement unit is a sensor integrating an accelerometer and a gyroscope, specifically employing MEMS devices, used to monitor robot posture changes in real time and compensate for motion errors. Gas sensors are devices that detect the concentration of specific gases. Specifically, they can use sensor modules based on electrochemical or infrared principles to monitor the concentration of oxygen, combustible gases, and toxic gases in cable trenches, preventing personnel from entering hazardous environments. It should be noted that, in some embodiments, when the mobile robot platform 1 autonomously moves within the cable trench, the LiDAR and inertial measurement unit continuously collect environmental data and construct a 3D map, while the binocular depth camera simultaneously acquires images and depth information. The multispectral illumination system automatically adjusts the supplementary lighting intensity according to the ambient light conditions. A high-precision odometer records the movement trajectory and fuses it with the LiDAR data to achieve accurate estimation of the robot's pose. The computational control module performs timestamp alignment and spatial coordinate binding on the multi-source data to generate structured data with spatial location information. An optional gas sensor monitors the gas concentration within the trench in real time, triggering an alarm signal when a dangerous threshold is detected and uploading it to a remote terminal via the communication module. All sensor data is processed by the computational control module and then transmitted to a cloud platform for storage and visualization via the communication module.

[0042] Through the above technical solutions, this application solves the safety hazards of manual auditing, avoiding personnel entering enclosed spaces lacking oxygen or containing harmful gases; it improves data collection efficiency, enabling full coverage inspection of cable trenches spanning several kilometers in a single inspection; through multi-sensor fusion and automated measurement, it reduces human error and improves the accuracy of construction element identification; through spatial coordinate binding and structured data management, it achieves precise location of defects and traceability of historical data; and through remote communication and cloud visualization, it supports multiple people collaboratively viewing audit results and quickly generating standardized reports.

[0043] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformations made based on the description and drawings of the present invention under the inventive concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A method for auditing cable trench construction based on robot imaging, characterized in that, Includes the following steps: The robot's audit path is pre-planned, and the robot moves autonomously along the cable trench; During the robot's movement, the robot uses data acquired by LiDAR and inertial measurement unit to construct map data of the cable trench and estimate the robot's pose data using SLAM algorithm; the robot collects image data and depth data through binocular cameras; and the robot uses a multispectral lighting system for ambient lighting; wherein the map data, pose data, image data, and depth data are bound to spatial position coordinates; Map data, pose data, image data, and depth data are aligned and fused in a timestamp and spatial coordinate system to generate a 3D point cloud model of the cable trench. The 3D point cloud model and image data of the cable trench are input into a pre-trained deep learning model. The deep learning model identifies and segments the construction elements in the cable trench, and generates a list of elements with unique IDs and location information in the spatial coordinate system for each construction element. Automated measurement is performed on each construction element in the element list to obtain measurement results, and the measurement results are compared with a pre-set construction specification standard database to diagnose defects that do not conform to the construction specification standards. Defect points are labeled in the 3D point cloud model of the cable trench, and a structured audit report is generated.

2. The method for auditing cable trench construction based on robot imaging according to claim 1, characterized in that, The specific steps for planning the robot's audit path are as follows: based on the cable trench design drawings, the robot's approximate audit path is planned in the background software; after the system is started, the steps also include completing the self-test of each sensor of the audit robot.

3. The method for auditing cable trench construction based on robot imaging according to claim 1, characterized in that, It also includes a lidar and an inertial measurement unit constructing a 3D map of the cable trench using a synchronous positioning and mapping algorithm, and estimating the robot pose using the synchronous positioning and mapping algorithm; and a binocular depth camera acquiring image data and depth data at specific intervals or continuously.

4. The method for auditing cable trench construction based on robot imaging according to claim 1, characterized in that, It also includes that the map data generated by the SLAM algorithm is based on a global coordinate system; the pose data is the robot's real-time position and orientation relative to the global coordinate system; the image data and depth data collected by the binocular camera are converted to the robot's body coordinate system through a pre-calibrated extrinsic parameter matrix, and uniformly registered to the global coordinate system according to the real-time pose data, so that the image data and its corresponding depth point cloud both have global coordinates.

5. The method for auditing cable trench construction based on robot imaging according to claim 1, characterized in that, It also includes the pre-trained deep learning model, which includes construction elements in the cable trench identified and segmented by a convolutional neural network, including cable supports, grounding flat irons, high-voltage cables, and fireproof partitions.

6. The method for auditing cable trench construction based on robot imaging according to claim 1, characterized in that, The automated measurement of construction elements includes: for cable supports, measuring the spacing between adjacent cable supports, the verticality of the cable support installation, and the flatness of the top surface of the cable support; for grounding flat irons, measuring the welding overlap length and bending radius of the grounding flat irons; and for high-voltage cables, detecting whether there is damage to the surface of the high-voltage cable and estimating the damaged area.

7. The method for auditing cable trench construction based on robot imaging according to claim 1, characterized in that, In the 3D point cloud model, the defect points are marked with highlights. The structured audit report includes an overview of the cable trench, an audit data overview, a defect list, and a link to browse the 3D point cloud model. The overview includes the trench location and audit time, the audit data overview includes the total length of the cable trench and the number of supports, and the defect list includes the defect type, defect location, defect image, and a comparison of the measured value with the standard value.

8. The method for auditing cable trench construction based on robot imaging according to claim 1, characterized in that, It also includes a cloud platform for storage; and enables the visualization of 3D models and audit results, specifically allowing users to remotely view the 3D point cloud model of the cable trench and the corresponding audit results through a web browser.

9. The method for auditing cable trench construction based on robot imaging according to claim 1, characterized in that, The robot is equipped with a gas sensor. During the robot's autonomous movement along the cable trench, the gas sensor is also used to detect the oxygen concentration, combustible gas concentration and toxic gas concentration in the cable trench. The data detected by the gas sensor is also bound to the spatial position coordinates and transmitted to the computing and control module.

10. A cable trench construction auditing device based on robot imagery, applied to the cable trench construction auditing method based on robot imagery as described in any one of claims 1-9, characterized in that, It includes a mobile robot platform, an integrated sensing module, a computing and control module, and a communication module; The integrated sensing module, the computing control module, and the communication module are all installed on the mobile robot platform, and the integrated sensing module and the communication module are electrically connected to the computing control module respectively. The integrated sensing module includes a binocular depth camera, lidar, high-precision odometer, multispectral illumination system, inertial measurement unit, and optional gas sensor. The binocular depth camera is located at the front end of the mobile robot platform, the lidar is located at the top of the mobile robot platform, the high-precision odometer is integrated into the mobile chassis of the mobile robot platform, the multispectral illumination system is arranged around the binocular depth camera, the inertial measurement unit is mounted on the mobile robot platform, and the optional gas sensor is mounted on the mobile robot platform. The binocular depth camera, lidar, high-precision odometer, multispectral illumination system, inertial measurement unit, and optional gas sensor are electrically connected to the computing and control module.

Images