Obstacle distance determination method and system

Abstract

The invention discloses an obstacle distance determination method and system, which are applied to the field of power system safety maintenance, and the method comprises the steps: obtaining infrared data, depth map data and environment data of a to-be-recognized region; determining a target obstacle; extracting related position information of a target obstacle in the depth map data to obtain three-dimensional point cloud data of the target obstacle; analyzing and processing the three-dimensional point cloud data to obtain the spatial main axis direction of the target obstacle and the temperature change condition of the target obstacle; processing the three-dimensional point cloud data, determining a geometric shape according to curvature distribution, and determining an initial risk distance according to the geometric shape; and based on the environment data and the temperature change condition, correcting the initial risk distance, and generating obstacle distance information. According to the obstacle distance determination method and system provided by the invention, objective, real-time and accurate quantification of the safe distance of the obstacle is realized, and the safety, reliability and intelligent level of electric power overhaul operation are improved.

Description

Technical Field

[0001] This invention relates to the field of power system safety maintenance technology, and in particular to a method and system for determining obstacle distance. Background Technology

[0002] Power maintenance is a key link in ensuring the safe and stable operation of the power grid. By regularly inspecting, maintaining, and promptly eliminating potential equipment hazards, we can not only effectively prevent power outages and equipment failures, but also extend the service life of power facilities and improve power supply reliability and power quality.

[0003] However, in actual operations, maintenance personnel often face various obstacles in the power line corridors, such as hanging debris and construction machinery. Accurately judging the spatial distance between obstacles and live equipment is crucial. Traditionally, this distance is mainly estimated visually or through experience. This method is significantly limited by factors such as obstructed viewpoints, light reflection, vertigo from heights, work fatigue, or distraction, making it easy for personnel to misjudge the actual distance. This can lead to unintentional intrusion into the safe clearance of live equipment, causing arcing or discharge. Furthermore, it can cause tools, personnel, or equipment to accidentally approach or touch high-voltage conductors, resulting in phase-to-phase short circuits or grounding faults, instantly damaging critical equipment such as circuit breakers and transformers, ultimately leading to widespread power outages and seriously threatening public safety and social order. Summary of the Invention

[0004] This invention provides a method and system for determining obstacle distances to solve the technical problem of inaccurate obstacle distance identification in existing technologies, thereby achieving accurate obstacle distance identification.

[0005] To address the aforementioned technical problems, this invention provides a method and system for determining obstacle distance, the method comprising: Acquire infrared data, depth map data, and environmental data of the area to be identified; The infrared data is processed to identify the target obstacle, and the target infrared data of the target obstacle is extracted from the infrared data. Extract the relevant location information of the target obstacle from the depth map data to obtain the three-dimensional point cloud data of the target obstacle; The three-dimensional point cloud data is analyzed and processed to obtain the spatial principal axis direction of the target obstacle, and the axial infrared data along the spatial principal axis direction in the target infrared data is analyzed to obtain the temperature change of the target obstacle. The three-dimensional point cloud data is processed to obtain the surface curvature of the target obstacle. The curvature distribution is used to determine the geometric shape, and the geometric shape is used to determine the initial risk distance. Based on the environmental data and temperature changes, the initial risk distance is corrected to generate obstacle distance information.

[0006] Preferably, the step of identifying the target obstacle by processing the infrared data and extracting the target infrared data of the target obstacle from the infrared data includes: The infrared data is preprocessed to reduce noise, resulting in smooth infrared data. An adaptive threshold segmentation technique is used to segment the smoothed infrared data to obtain candidate regions; The candidate regions are filtered based on their areas to obtain candidate obstacle region data; The candidate obstacle region data is input into a preset infrared feature matching algorithm to determine the target obstacle; The infrared data of the target obstacle is obtained by extracting the infrared data within the pixel coordinate range corresponding to the target obstacle from the infrared data.

[0007] Preferably, the step of analyzing the axial infrared data along the spatial principal axis in the target infrared data to obtain the temperature change of the target obstacle includes: The pixel coordinate matrix and the corresponding temperature value matrix of the target infrared data are extracted to obtain a structured infrared temperature dataset. Based on the structured infrared temperature dataset and the spatial principal axis direction, the spatial principal axis direction is transformed into a projection line equation in the pixel coordinate system through coordinate transformation, thereby obtaining the pixel projection path of the spatial principal axis direction in the target infrared data; Temperature data is extracted from the target infrared data based on the pixel projection path using an interpolation algorithm to obtain temperature sequence data; The temperature sequence data is processed to obtain the temperature change information.

[0008] Preferably, the step of processing the three-dimensional point cloud data to obtain the surface curvature of the target obstacle, determining the geometric shape based on the curvature distribution, and determining the initial risk distance based on the geometric shape includes: The 3D point cloud data is denoised to obtain denoised 3D point cloud data. For each target point in the denoised 3D point cloud data, a fixed number of neighboring points are selected, and the local plane equation of the neighboring points is fitted to obtain the local curvature data of each target point. The local curvature data of each target point is analyzed and clustered to obtain the curvature distribution of the target obstacle. The curvature distribution is analyzed to determine the geometry of the target obstacle; The geometric shape is processed according to a preset initial risk distance rule base for power scenarios to obtain the initial risk distance.

[0009] Preferably, the step of correcting the initial risk distance based on the environmental data and the temperature changes to generate obstacle distance information for the area to be identified includes: The environmental data and temperature changes are coupled and analyzed to determine the correction coefficient for the distance to the target obstacle; The initial risk distance is corrected based on the correction coefficient, and the obstacle distance information is generated from the correction result.

[0010] Another aspect of the present invention provides a system for determining obstacle distance, comprising: The acquisition module is used to acquire infrared data, depth map data, and environmental data of the area to be identified. The identification module is used to identify the infrared data, determine the target obstacle, and extract the target infrared data of the target obstacle from the infrared data. The extraction module is used to extract the relevant location information of the target obstacle from the depth map data to obtain the three-dimensional point cloud data of the target obstacle; The analysis module is used to analyze and process the three-dimensional point cloud data to obtain the spatial principal axis direction of the target obstacle, and to analyze the axial infrared data along the spatial principal axis direction in the target infrared data to obtain the temperature change of the target obstacle. The processing module is used to process the three-dimensional point cloud data to obtain the surface curvature of the target obstacle, determine the geometric shape based on the curvature distribution, and determine the initial risk distance based on the geometric shape. The generation module is used to correct the initial risk distance based on the environmental data and the temperature changes, and generate obstacle distance information.

[0011] Preferably, the identification module includes: A noise reduction unit is used to perform noise reduction preprocessing on the infrared data to obtain smooth infrared data; The segmentation unit is used to segment the smooth infrared data using an adaptive threshold segmentation technique to obtain candidate regions; A filtering unit is used to filter the candidate regions based on their area to obtain candidate obstacle region data; The matching algorithm unit is used to input the candidate obstacle region data into a preset infrared feature matching algorithm to determine the target obstacle; An extraction unit is used to extract infrared data within the pixel coordinate range corresponding to the target obstacle from the infrared data to obtain the target infrared data.

[0012] Preferably, the step of analyzing the axial infrared data along the spatial principal axis in the target infrared data to obtain the temperature change of the target obstacle includes: The structured unit is used to extract the pixel coordinate matrix and the corresponding temperature value matrix from the target infrared data to obtain a structured infrared temperature dataset. The projection path unit is used to transform the spatial principal axis direction into a projection line equation in the pixel coordinate system based on the structured infrared temperature dataset and the spatial principal axis direction, thereby obtaining the pixel projection path of the spatial principal axis direction in the target infrared data. The extraction unit is used to extract temperature data from the target infrared data based on the pixel projection path using an interpolation algorithm to obtain temperature sequence data; The processing unit is used to process the temperature sequence data to obtain the temperature change information.

[0013] Preferably, the processing module includes: A denoising unit is used to denoise the three-dimensional point cloud data to obtain denoised three-dimensional point cloud data. A local curvature unit is used to select a fixed number of neighboring points for each target point in the denoised 3D point cloud data, fit the local plane equation of the neighboring points, and obtain the local curvature data of each target point. A clustering unit is used to analyze and cluster the local curvature data of each target point to obtain the curvature distribution of the target obstacle. A curvature unit is used to analyze the curvature distribution and determine the geometry of the target obstacle. The preset unit is used to process the geometric shape according to the preset initial risk distance rule library of power scenario to obtain the initial risk distance.

[0014] Preferably, the generation module includes: A coupling unit is used to perform coupling analysis on the environmental data and the temperature change to determine the correction coefficient for the distance to the target obstacle; The correction unit is used to correct the initial risk distance based on the correction coefficient, and generate the obstacle distance information based on the correction result.

[0015] Compared with the prior art, the beneficial effects of the present invention are at least one of the following: This invention constructs a 3D point cloud of a target obstacle by fusing infrared data, depth map data, and environmental data, and extracts its spatial principal axis direction. It then analyzes the axial infrared temperature variation along the principal axis to characterize the electrical state. Simultaneously, it determines the obstacle's geometry based on the point cloud surface curvature, and sets an initial risk distance that conforms to the electric field distribution characteristics. Furthermore, by combining environmental data and the smoothness of temperature changes, it dynamically corrects the initial risk distance in a physically interpretable manner, generating obstacle distance information that includes minimum and maximum safe distances. This method abandons the traditional approach of relying on visual estimation or experience, achieving objective, real-time, and accurate quantification of obstacle safety distances. It effectively avoids distance misjudgments caused by viewpoint obstruction, light interference, or personnel fatigue, significantly reducing the risk of discharge, short circuits, or equipment damage caused by accidentally entering the safe clearance distance, thereby improving the safety, reliability, and intelligence level of power maintenance operations. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a method for determining obstacle distance in one embodiment of the present invention; Figure 2 This is a schematic diagram of the obstacle distance determination system in one embodiment of the present invention; Figure label: The module comprises: 11. Acquisition module; 12. Identification module; 13. Extraction module; 14. Analysis module; 15. Processing module; and 16. Generation module. Detailed Implementation

[0017] 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 some embodiments of the present invention, and not all embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. 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.

[0018] In the description of this application, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0019] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "joint" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. The term "and / or" as used herein includes any and all combinations of one or more of the related listed items. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0020] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0021] One embodiment of the present invention provides a method for determining the distance to an obstacle; for details, please refer to [link to documentation]. Figure 1 , Figure 1 The diagram shown is a flowchart illustrating a method for determining obstacle distance in one embodiment of the present invention, including: S1. Acquire infrared data, depth map data, and environmental data of the area to be identified; S2. Perform identification processing on the infrared data to determine the target obstacle and extract the target infrared data of the target obstacle from the infrared data; S3. Extract the relevant location information of the target obstacle from the depth map data to obtain the three-dimensional point cloud data of the target obstacle; S4. Analyze and process the three-dimensional point cloud data to obtain the spatial principal axis direction of the target obstacle, and analyze the axial infrared data along the spatial principal axis direction in the target infrared data to obtain the temperature change of the target obstacle. S5. Process the three-dimensional point cloud data to obtain the surface curvature of the target obstacle, determine the geometric shape based on the curvature distribution, and determine the initial risk distance based on the geometric shape. S6. Based on environmental data and temperature changes, the initial risk distance is corrected to generate obstacle distance information.

[0022] Preferably, infrared data, depth map data, and environmental data of the area to be identified are acquired. Infrared data refers to the infrared radiation signal data generated by objects within the area due to temperature differences, captured by infrared thermal imaging equipment. This data reflects the temperature distribution characteristics of the objects. Depth map data is three-dimensional spatial data representing the distance between each point in the area and the equipment, acquired using devices such as depth cameras or lidar. Environmental data includes external environmental parameters affecting signal propagation and equipment perception, such as temperature, humidity, air pressure, light intensity, and air composition. Acquiring these three types of data comprehensively characterizes the state of the area to be identified from three dimensions: temperature characteristics, spatial location, and external interference factors, providing complete data support for subsequent target obstacle identification and distance judgment. Specifically, a mobile detection device integrating an infrared thermal imaging module, lidar, and multi-parameter environmental sensors can be used. Infrared data, depth map data, and environmental data are acquired through fixed-point collection or mobile scanning at the power maintenance site. The advantage of this approach is that it can simultaneously acquire multi-dimensional features of the target, avoiding the perception limitations of single data types in complex scenarios and improving the accuracy and reliability of subsequent target identification and distance calculation.

[0023] Next, the infrared data is processed to identify target obstacles and extract their infrared data. Noise reduction preprocessing is performed on the infrared data to obtain smoothed infrared data. Adaptive threshold segmentation technology is used to segment the smoothed infrared data to obtain candidate regions. These candidate regions are then filtered based on their area to obtain candidate obstacle region data. This candidate obstacle region data is input into a pre-defined infrared feature matching algorithm to determine the target obstacle. Finally, infrared data within the pixel coordinate range corresponding to the target obstacle is extracted from the infrared data to obtain the target infrared data.

[0024] Noise reduction preprocessing is the process of removing random noise and environmental interference from infrared data using techniques such as filtering; smoothed infrared data refers to infrared data with clear edges and stable temperature information after noise reduction preprocessing; adaptive threshold segmentation is an image segmentation technique that automatically adjusts the segmentation threshold based on the gray value distribution of local image regions to separate the target region from the background; candidate region refers to the region that may contain obstacles, separated from the smoothed infrared data by adaptive threshold segmentation; the area of ​​the candidate region refers to the total number of pixels contained in the candidate region; filtering is the process of setting a reasonable range to eliminate regions that do not meet the requirements; candidate obstacle region data refers to... Candidate region data that matches the actual obstacle size characteristics after area screening; the preset infrared feature matching algorithm refers to the algorithm that identifies targets by comparing the features of candidate regions with the feature library templates based on a pre-built infrared feature library of common obstacles in power maintenance scenarios; target obstacles refer to objects in power maintenance scenarios that may threaten the safety of live equipment, such as hanging foreign objects and construction machinery, as confirmed by the infrared feature matching algorithm; pixel coordinate range refers to the pixel position boundary occupied by the target obstacle in the infrared data image; target infrared data refers to the temperature value and radiation intensity and other related data extracted from the infrared data within the pixel coordinate range corresponding to the target obstacle.

[0025] The core of infrared data recognition and processing is to accurately locate target obstacles and extract their unique infrared information. Infrared data is processed to identify and extract target infrared data. First, infrared data noise reduction preprocessing is performed. In practice, a 5×5 Gaussian filtering method is used, calculating a weighted average of each pixel and its neighboring pixels using a Gaussian function with a standard deviation of 1.2. Alternatively, a 3×3 window median filter is used, sorting the grayscale values ​​of pixels within the window and replacing the center pixel with the median value. This removes isolated noise points caused by device noise and environmental interference, resulting in smooth infrared data with continuous edges and accurate temperature information. Next, adaptive threshold segmentation technology is employed, specifically using an adaptive threshold algorithm based on local means. The smooth infrared data is divided into local regions of 16×16 pixels, and the mean and standard deviation of each local region are calculated. Then, the threshold is calculated using the formula: "Threshold = Local Mean - 0.8 × Local Standard Deviation". The system automatically generates segmentation thresholds for each region, classifying pixels with grayscale values ​​above the threshold as foreground regions and those below as background regions, thus obtaining candidate regions that may contain obstacles. Then, it performs a filtering process based on the area of ​​these candidate regions. Considering the actual dimensions of common obstacles in power maintenance scenarios, such as hanging debris and construction machinery, the area filtering range is set to 50 to 5000 pixels. An image pixel counting tool is used to count the number of pixels in each candidate region, eliminating small noise regions with fewer than 50 pixels and large background regions with more than 5000 pixels, resulting in candidate obstacle region data. This candidate obstacle region data is then input into a pre-built infrared feature matching algorithm. This algorithm is based on a pre-constructed feature library, which includes the average temperature range of hanging debris such as plastic sheeting. The system analyzes the temperature distribution characteristics of obstacles within a 15-25°C range and the average temperature range (30-45°C) and shape contour characteristics of metal components in construction machinery such as cranes. It calculates the temperature mean radiation intensity histogram and edge contour moments of candidate obstacle regions and compares them with the feature parameters of various obstacles in the feature library using Euclidean distance. When the similarity exceeds 90%, the obstacle is identified as a target obstacle. Finally, the pixel coordinate range of the target obstacle is obtained from the infrared data using image coordinate positioning tools. The infrared radiation values ​​and calculated temperature values ​​of all pixels within this range are extracted to obtain the target infrared data. This approach, with its specific and implementable technical parameters and operational steps, ensures precise execution at each step, improving the accuracy of target obstacle identification and providing a high-quality data foundation for subsequent temperature analysis. This multi-step process gradually narrows down the target range, improving the accuracy of target obstacle identification, while the extracted target infrared data provides precise data support for subsequent analysis of temperature changes.

[0026] Furthermore, the relevant location information of the target obstacle in the depth map data is extracted to obtain the 3D point cloud data of the target obstacle. The depth map data is a two-dimensional image acquired by devices such as depth cameras or LiDAR, representing the actual distance between each pixel in the area to be identified and the device. Each pixel value corresponds to the vertical distance from that point to the device. The relevant location information refers to the coordinate parameters and spatial distribution characteristics of the target obstacle in 3D space. The 3D point cloud data is a dataset composed of a large number of discrete 3D coordinate points that can intuitively reflect the spatial shape of the object. Each point contains coordinate values ​​in three directions: X-axis, Y-axis, and Z-axis. First, the pixel coordinate range of the determined target obstacle in the infrared data is mapped to the depth map data using image registration technology. Image registration technology can adopt a feature point-based matching method. SIFT (Scale-Invariant Feature Transform) feature points or ORB (Oriented Fast and Rotated) feature points are extracted from the infrared image and depth map respectively. BRIEF (directional FAST and rotated BRIEF) feature points are matched by calculating the Euclidean distance between feature points, and then the transformation matrix is ​​solved based on the matched feature point pairs to complete the accurate mapping of pixel coordinates, obtaining the pixel region of the target obstacle in the depth map. Next, the depth value extraction tool reads the depth value corresponding to each pixel in the pixel region, and combines the intrinsic parameters of the depth camera or LiDAR, including focal length, principal point coordinates, and distortion coefficients, to convert the two-dimensional coordinates and depth values ​​of each pixel into three-dimensional spatial coordinates using the camera imaging model formula. Finally, all the converted three-dimensional coordinate points are summarized to form three-dimensional point cloud data that can completely present the spatial shape and position of the target obstacle. This is done to transform two-dimensional depth information into three-dimensional spatial data, providing a basis for subsequent analysis of the spatial principal axis direction and surface curvature of the target obstacle. The advantage is that it can accurately restore the three-dimensional shape of the target obstacle, avoid the limitation that two-dimensional data cannot reflect spatial positional relationships, and improve the accuracy of subsequent processing.

[0027] Preferably, the 3D point cloud data is analyzed and processed to obtain the spatial principal axis direction of the target obstacle, and the axial infrared data along the spatial principal axis direction in the target infrared data is analyzed to obtain the temperature change of the target obstacle. The pixel coordinate matrix and the corresponding temperature value matrix are extracted from the target infrared data to obtain a structured infrared temperature dataset. Based on the structured infrared temperature dataset and the spatial principal axis direction, the spatial principal axis direction is transformed into a projection line equation in the pixel coordinate system through coordinate transformation to obtain the pixel projection path of the spatial principal axis direction in the target infrared data. Based on the pixel projection path, an interpolation algorithm is used to extract temperature data from the target infrared data to obtain temperature sequence data. The temperature sequence data is processed to obtain the temperature change.

[0028] The 3D point cloud data is analyzed to obtain the spatial principal axis direction of the target obstacle, and the temperature change is analyzed in conjunction with the target infrared data. The spatial principal axis direction refers to the 3D spatial axis direction that reflects the overall extension trend of the target obstacle, such as the length extension direction of a long, hanging object or the symmetrical axis direction of a construction machine. The structured infrared temperature dataset is a matrix formed by arranging the pixel coordinates of the target infrared data in rows and columns, combined with a matrix composed of the temperature values ​​corresponding to each pixel coordinate, forming a standardized data set containing both positional and temperature information. The projection line equation in the pixel coordinate system maps the 3D spatial principal axis direction through mathematical transformation. After reaching the pixel plane of the 2D infrared image, the linear mathematical expression describing the projection trajectory is used. The pixel projection path refers to the sequence of pixels formed by projecting the principal axis of space onto the infrared image. The interpolation algorithm is a mathematical method to calculate the temperature at a location when the pixel projection path passes through non-integer pixel coordinates, using the known temperature values ​​of surrounding pixels. The temperature sequence data is a set of sequentially arranged temperature values ​​extracted along the pixel projection path. Temperature variation refers to the fluctuation range, trend, and extreme value distribution of temperature values ​​in the temperature sequence data. In specific implementation, the 3D point cloud data is first processed using principal component analysis to calculate the covariance matrix of the point cloud data, and then the eigenvalues ​​of the covariance matrix are solved. The direction corresponding to the eigenvector with the largest eigenvalue is the spatial principal axis direction of the target obstacle. Next, a pixel coordinate matrix is ​​extracted from the target infrared data. This matrix uses the row and column numbers of pixels in the image as elements. Simultaneously, the temperature value corresponding to each pixel is extracted to form a temperature value matrix. The two are combined to form a structured infrared temperature dataset. Then, based on the three-dimensional vector of the spatial principal axis direction and the intrinsic parameters of the infrared camera, the three-dimensional vector is converted into a direction vector in the pixel coordinate system using the perspective projection transformation formula. Combined with the pixel coordinates of the center point of the target obstacle in the infrared image, this is substituted into the linear equation y = kx + b (where k is the slope of the direction vector). (b is the intercept), to obtain the pixel projection path of the spatial principal axis in the target infrared data; then, based on the pixel projection path, a bilinear interpolation algorithm is used. When the projection path passes through the pixel (u+Δu, v+Δv) (where u and v are integer pixel coordinates, and Δu and Δv are decimals from 0 to 1), the temperature values ​​of the four integer coordinate pixels (u, v), (u+1, v), (u, v+1) and (u+1, v+1) around the pixel are taken, and the temperature value at that position is calculated according to the weights (1-Δu)×(1-Δv), Δu×(1-Δv), (1-Δu)×Δv and Δu×Δv. The temperature sequence data is obtained by continuously calculating along the projection path.Finally, a sliding window averaging method was used to remove random fluctuations in the temperature sequence data. The fluctuation range was obtained by calculating the maximum, minimum, and temperature differences. The trend was then obtained through linear fitting or trend line analysis, thereby determining the temperature changes of the target obstacle. This approach was taken to accurately analyze the temperature distribution along the main extension direction of the obstacle, avoiding temperature analysis bias caused by random direction selection. It combines the spatial morphology and temperature characteristics of the obstacle to provide a more realistic temperature basis for subsequent risk distance correction.

[0029] Preferably, the 3D point cloud data is processed to obtain the surface curvature of the target obstacle. The curvature distribution is used to determine the geometric shape, and the geometric shape is used to determine the initial risk distance. The 3D point cloud data is denoised to obtain denoised 3D point cloud data. A fixed number of neighboring points are selected for each target point in the denoised 3D point cloud data, and the local plane equations of the neighboring points are fitted to obtain the local curvature data of each target point. The local curvature data of each target point is analyzed and clustered to obtain the curvature distribution of the target obstacle. The curvature distribution is analyzed to determine the geometric shape of the target obstacle. The geometric shape is processed according to a preset initial risk distance rule base for power scenarios to obtain the initial risk distance.

[0030] 3D point cloud data is processed to obtain the surface curvature of the target obstacle, thereby determining its geometric shape and initial risk distance. Surface curvature is a physical quantity describing the degree of bending of the target obstacle's surface, and is divided into Gaussian curvature and mean curvature. Gaussian curvature reflects the overall bending trend of the surface, while mean curvature reflects the local unevenness of the surface. Denoising 3D point cloud data is data obtained by filtering out anomalies caused by equipment errors or environmental interference from the original 3D point cloud data. Target points refer to the discrete points in the denoised 3D point cloud data that constitute the surface of the target obstacle. Neighborhood points are surrounding points selected around each target point to calculate the local geometric features of that target point. The local plane equation is obtained by fitting the neighborhood points and describes... The mathematical equation for the local surface trend near the target point; the local curvature data is the curvature value of each target point calculated based on the local plane equation; the curvature distribution is the distribution characteristic formed by classifying and statistically analyzing the local curvature data of all target points according to their numerical range; geometric morphology refers to the overall shape characteristics of the target obstacle, such as spherical, cylindrical, or irregular sheet-like shapes; the preset initial risk distance rule base for power scenarios is a database built based on power safety specifications and historical cases, storing safety distance standards between obstacles of different geometric shapes and live equipment; the initial risk distance is a preliminary safety distance value obtained by matching the obstacle's geometric shape from the rule base; in specific implementation, the 3D point cloud data is first denoised using statistical filtering, and each... The number of neighboring points for each point is 20. The mean and standard deviation of the distance between each point and its neighbors are calculated. Points with a distance greater than the mean plus twice the standard deviation are identified as outliers and removed, resulting in denoised 3D point cloud data. Next, for each denoised target point, 15 neighboring points are selected. The local plane equation of the neighboring points is fitted using the least squares method. Assuming the plane equation is ax + by + cz + d = 0, the 3D coordinates of the neighboring points are substituted to establish a system of equations. The coefficients a, b, c, and d are solved, and the Gaussian curvature and mean curvature of the target point are calculated using differential geometry formulas to obtain local curvature data. Then, the K-means clustering algorithm is used to cluster the local curvature data of all target points, with the number of clusters set to 3 to 5. The curvature data is assigned to a category by calculating the Euclidean distance between the curvature data and the cluster centers. The distribution range of curvature values ​​and the number of corresponding points for each category are statistically analyzed to obtain the curvature distribution. Then, the curvature distribution is analyzed. If the Gaussian curvature of most points is close to 0 and the absolute value of the average curvature is small, the geometric shape is determined to be cylindrical. If there are a large number of points with positive Gaussian curvature and similar average curvature, the shape is determined to be spherical. If the curvature value distribution is chaotic and has no obvious pattern, the shape is determined to be irregular sheet-like. Finally, the preset initial risk distance rule library for the power scenario is called. The rule library stores that the initial risk distance for cylindrical obstacles such as steel pipes is 1.5 meters, the initial risk distance for spherical obstacles such as plastic balls is 1 meter, and the initial risk distance for irregular sheet-like obstacles such as plastic sheeting is 0.An initial risk distance of 8 meters is obtained by matching the determined geometric shape with the corresponding numerical value. This is done to establish a preliminary safe distance based on the actual geometric characteristics of the obstacle, avoiding the irrationality of a uniform distance standard. The advantage is that it allows the initial risk distance to accurately match the actual shape of the obstacle, providing a scientific basis for subsequent adjustments.

[0031] Finally, based on environmental data and temperature changes, the initial risk distance is corrected to generate obstacle distance information. A coupled analysis of environmental data and temperature changes is performed to determine the correction coefficient for the target obstacle distance; the initial risk distance is then corrected based on this coefficient, and the corrected result is used to generate obstacle distance information.

[0032] The initial risk distance is corrected based on environmental data and temperature changes to generate obstacle distance information. The coupling analysis refers to combining two mutually influential factors, environmental data and temperature changes, to analyze their combined effect on the obstacle safety distance. The correction coefficient is a proportional coefficient used to adjust the initial risk distance, calculated based on the coupling analysis results. This coefficient is greater than 0 and is usually between 0.8 and 1.2. A value greater than 1 indicates that the distance needs to be increased, and a value less than 1 indicates that the distance needs to be decreased. The obstacle distance information is the final data obtained after correction, which can accurately reflect the safety distance between the target obstacle and the energized equipment. First, the key parameters included in the environmental data are identified, such as air humidity, ambient temperature, air pressure, and light intensity. Temperature variations include temperature fluctuation range and extreme temperatures. Next, a coupled analysis model is constructed, using a weighted summation algorithm to quantify the impact of each parameter on distance. For example, the weights are set as follows: air humidity 0.3, ambient temperature 0.2, air pressure 0.2, and temperature fluctuation range 0.3. Each parameter is then standardized. For instance, with air humidity as a baseline of 60%, the standardized value is (actual humidity - 60%) / 40% when it is above 60%, and below 60%... (60% - actual humidity) / 60%, ambient temperature is based on 25℃, the standardized value is (actual temperature - 25℃) / 15℃ when above 25℃, and (25℃ - actual temperature) / 25℃ when below 25℃, air pressure is based on standard atmospheric pressure 101.3 kPa, the standardized value is (actual air pressure - 101.3) / 5, temperature fluctuation range is based on 5℃, the standardized value is (actual fluctuation range - 5) / 5; then, the comprehensive impact value is calculated based on the standardized parameter values ​​and corresponding weights. If the comprehensive impact value is positive, it indicates that environmental and temperature factors will increase. Adding safety risk, the correction coefficient is 1 + 0.2 × comprehensive impact value. If the comprehensive impact value is negative, it indicates that the factor has a small impact on the risk, and the correction coefficient is 1 + 0.1 × comprehensive impact value. For example, when the air humidity is 80% (standardized value 0.5), the ambient temperature is 30℃ (standardized value 0.33), the air pressure is 103kPa (standardized value 0.34), and the temperature fluctuation range is 8℃ (standardized value 0.6), the comprehensive impact value = 0.5 × 0.3 + 0.33 × 0.2 + 0.34 × 0.2 + 0.6 × 0.3 = 0.15 + 0.066 + 0.0 68 + 0.18 = 0.464, correction factor = 1 + 0.2 × 0.464 = 1.0928; then multiply the correction factor by the initial risk distance. For example, if the initial risk distance is 1 meter, the corrected distance = 1 × 1.0928 ≈ 1.09 meters. Finally, integrate the corrected distance value with the corresponding ambient temperature conditions to generate obstacle distance information. This is done to ensure that the safety distance takes into account the actual environment and temperature effects, avoiding the initial distance from being out of sync with the on-site situation. The benefit is that it can improve the accuracy of obstacle distance information and provide a more reliable basis for safe operation of power maintenance.

[0033] Another embodiment of the present invention provides a live-line work obstacle avoidance monitoring system based on multi-sensor fusion. The system mainly includes a lidar perception module, a depth vision module, a data processing and control center, a data fusion and early warning module, and a human-machine interface. The lidar perception module is responsible for three-dimensional scanning of the working environment, acquiring high-precision real-time point cloud data by emitting laser beams and receiving echoes. The core function of this module is to process the point cloud data, including filtering and denoising the point cloud, clustering and segmenting to separate independent obstacle targets, and then extracting the geometric features of these targets (such as size, shape, and contour), and identifying obstacle types (such as personnel, vehicles, and power equipment) based on a predefined feature library or machine learning model. More importantly, this module needs to continuously track the identified obstacles, predict their trajectories using algorithms (such as Kalman filtering), and calculate in real time the three-dimensional spatial distance between them and preset safety points (such as the position of a worker's hand or the basket of an insulated bucket truck). The depth vision module works in conjunction with the lidar module, and its core equipment is a depth camera (such as a structured light camera, a binocular stereo camera, or a time-of-flight camera). This module can directly capture video streams containing depth information for each pixel. Through frame-by-frame analysis of the video stream, it uses a deep learning-based target detection algorithm to quickly and accurately identify multiple targets in the image. Subsequently, a multi-target tracking algorithm is applied to assign a unique ID to each detected target and maintain its trajectory across consecutive frames. Based on the depth information provided by the depth camera itself, the actual physical distance between any two points in the image can be directly calculated, or the distance between the tracked target and a specific reference point in the image (such as the boundary of the work area) can be calculated. The data processing and control center is the brain of the system, typically powered by a high-performance embedded processor or industrial computer, equipped with sufficient computing resources and memory. The data processing and control center connects to the LiDAR sensing module and the depth vision module via a high-speed data interface, constructing an efficient multimodal data pipeline. This center first parses and preprocesses the received raw sensor data, with a key innovation being the introduction of a dynamic adaptive spatiotemporal registration mechanism. This mechanism not only performs a rigid transformation from the lidar point cloud coordinate system and the depth camera image coordinate system to a unified world coordinate system, but also compensates for minute sensor pose drifts caused by platform vibration or temperature changes in real time through online calibration technology, ensuring precise alignment of multi-source data in the spatiotemporal dimensions. At the data association and fusion level, the system employs a novel hierarchical collaborative fusion strategy. This strategy deeply integrates the rich textural semantic information of vision with the precise geometric information of lidar. Through a special cross-modal attention mechanism, it performs weighted fusion of the observation features of the same target in different sensors. For example, it utilizes the detailed image features provided by depth vision to assist lidar in identifying and classifying small obstacles.For successfully associated targets, the system does not simply use fixed-weight Kalman filtering, but proposes and employs a confidence-based adaptive fusion algorithm. This algorithm dynamically evaluates the LiDAR point cloud density, depth vision image clarity, and target motion state, adjusting the fusion weights of different sensor data in real time to achieve the optimal balance between target tracking smoothness and response speed. For targets that fail to be successfully associated in the current frame, the system initiates a multi-hypothesis tracking mechanism, establishing a temporary trajectory and performing multi-frame confidence assessments, rather than simply retaining or discarding them, effectively reducing false alarms and missed alarms. Finally, the system generates a high-precision, real-time updated comprehensive environmental situation map. This situation map not only includes the precise 3D position, velocity vector, unique ID tag, and type identified based on multimodal information for all static and dynamic obstacles in the environment, but also integrates scene depth completion information derived from depth completion technology for the first time, effectively compensating for the LiDAR's perception blind spots in extremely close areas, making the perception of small obstacles and complex structures more complete. Based on this comprehensive environmental situation map, the data fusion and early warning module executes core security decisions. The module's internal preset safety threshold library is not fixed but can be fine-tuned based on a dynamic spatial weighting model. This model can dynamically adjust the safety margin for different areas and obstacle types based on the identified task stage and obstacle type. The system calculates the distances between targets in the situation map in real time, especially the critical distances between workers, equipment, and live conductors. At the decision-making level, the module adopts a safety margin-oriented fusion decision-making mechanism with a conservative decision logic: when distance information from different sensors differs, the system will adopt the smaller effective distance value as the basis for risk judgment based on the worst-case analysis principle, thus building a redundant safety defense line at the sensor level. Once any distance calculation result is less than the dynamically adjusted safety threshold, the module will immediately trigger a graded early warning mechanism to ensure operational safety. Early warning signals can be graded according to risk level (such as warning, alarm, emergency alarm) and sent to workers and managers through various means such as visual (such as interface flashing, red light), audible (such as buzzers of different frequencies), or tactile (such as vibration of wearable devices), achieving proactive and precise risk intervention. The human-machine interface (HMI) serves as a bridge between the system and the user, typically presented as a graphical interface on tablets, in-vehicle displays, or augmented reality glasses. This interface renders and displays a comprehensive environmental situational map in real-time, marking obstacles with information such as their location, type, and distance using 3D models or augmented reality tags. Simultaneously, warning information is highlighted on this interface, and it may provide suggested obstacle avoidance guidance or operational prompts, greatly enhancing the operator's environmental awareness and situational awareness. Furthermore, the system can be expanded to integrate a maintenance resource optimization and scheduling module.This module utilizes precise environmental data (such as the location of faulty equipment and the complexity of the surrounding environment) and equipment information generated by the system, combined with the maintenance resource status in the database (such as team location, skills, and tool availability), and employs path planning and scheduling optimization algorithms to provide the command center with the optimal maintenance resource dispatch plan, thereby improving overall operation and maintenance efficiency. Corresponding to the above system, this invention also provides a method for obstacle avoidance monitoring in live-line work based on multi-sensor fusion. This method mainly includes the following steps: simultaneously collecting environmental data through lidar and depth vision sensors; preprocessing, detecting, identifying, and tracking point cloud data and depth video streams respectively; fusing multi-source sensing data to generate a unified three-dimensional environmental situation map; calculating key distances in real time based on the situation map and comparing them with safety thresholds; and initiating a graded early warning procedure when a potential collision risk is detected. The method also includes extended steps for maintenance decision support using situation map data.

[0034] Another embodiment of the present invention provides a system for determining obstacle distance; for details, please refer to [link to relevant documentation]. Figure 2 , Figure 2 The diagram shown illustrates a structural block diagram of an obstacle distance determination system according to one embodiment of the present invention, comprising: The acquisition module 11 is used to acquire infrared data, depth map data and environmental data of the area to be identified; The identification module 12 is used to identify and process infrared data, determine target obstacles, and extract target infrared data of target obstacles from the infrared data. Extraction module 13 is used to extract the relevant location information of the target obstacle in the depth map data to obtain the three-dimensional point cloud data of the target obstacle; Analysis module 14 is used to analyze and process three-dimensional point cloud data to obtain the spatial principal axis direction of the target obstacle, and to analyze the axial infrared data along the spatial principal axis direction in the target infrared data to obtain the temperature change of the target obstacle. Processing module 15 is used to process three-dimensional point cloud data to obtain the surface curvature of the target obstacle, determine the geometric shape based on the curvature distribution, and determine the initial risk distance based on the geometric shape. The generation module 16 is used to correct the initial risk distance based on environmental data and temperature changes, and generate obstacle distance information.

[0035] Preferably, the identification module 12 includes: The noise reduction unit is used to perform noise reduction preprocessing on the infrared data to obtain smooth infrared data; The segmentation unit is used to segment smooth infrared data using adaptive threshold segmentation technology to obtain candidate regions; The filtering unit is used to filter candidate regions based on their area to obtain candidate obstacle region data; The matching algorithm unit is used to input candidate obstacle region data into a preset infrared feature matching algorithm to determine the target obstacle; The extraction unit is used to extract infrared data within the pixel coordinate range corresponding to the target obstacle from the infrared data, thereby obtaining the target infrared data.

[0036] Preferably, the axial infrared data along the principal axis of space in the target infrared data is analyzed to obtain the temperature change of the target obstacle, including: The structured unit is used to extract the pixel coordinate matrix and the corresponding temperature value matrix from the target infrared data to obtain the structured infrared temperature dataset. The projection path unit is used to transform the spatial principal axis direction into the projection line equation in the pixel coordinate system based on the structured infrared temperature dataset and the spatial principal axis direction, so as to obtain the pixel projection path of the spatial principal axis direction in the target infrared data. The extraction unit is used to extract temperature data from the target infrared data based on the pixel projection path using an interpolation algorithm to obtain temperature sequence data; The processing unit is used to process the temperature sequence data to obtain the temperature change information.

[0037] Preferably, the processing module 15 includes: The denoising unit is used to denoise the 3D point cloud data to obtain denoised 3D point cloud data. The local curvature unit is used to select a fixed number of neighborhood points for each target point in the denoised 3D point cloud data, fit the local plane equation of the neighborhood points, and obtain the local curvature data of each target point. Clustering units are used to analyze and cluster the local curvature data of each target point to obtain the curvature distribution of the target obstacle; Curvature elements are used to analyze curvature distribution and determine the geometry of target obstacles. The preset unit is used to process the geometric shape according to the preset initial risk distance rule library of the power scenario to obtain the initial risk distance.

[0038] Preferably, the generation module 16 includes: The coupling unit is used to perform coupled analysis of environmental data and temperature changes to determine the correction coefficient for the distance to the target obstacle; The correction unit is used to correct the initial risk distance based on the correction coefficient, and generate obstacle distance information based on the correction result.

[0039] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0040] Accordingly, embodiments of the present invention provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform steps in the obstacle distance determination method of the above embodiments, for example... Figure 1 Steps S1 to S6 as described above.

[0041] This invention, by fusing infrared data, depth map data, and environmental data, not only achieves joint perception of the spatial location and electrical state of obstacles but also further explores the physical correlations between multimodal information. Specifically, it utilizes the spatial gradient characteristics of the infrared temperature field to identify potential hot spots and accurately reconstructs the three-dimensional relative pose between obstacles and charged conductors using depth point clouds. Based on this, it dynamically models the air insulation strength according to environmental parameters such as temperature, humidity, air pressure, and wind speed, and introduces the rate of temperature change as an indicator of equipment thermal stability to adaptively adjust the safety distance threshold. This method overcomes the limitations of the traditional fixed safety distance "one-size-fits-all" approach, constructing a risk boundary system closely coupled with real-time operating conditions. It can provide early warning of the risk of reduced clearance due to thermal expansion or wind deflection, and avoid efficiency losses caused by overly conservative operations in high insulation margin environments. Therefore, while ensuring the safety of personnel and equipment, it significantly improves the scientific nature of maintenance strategies, the timeliness of response, and the accuracy of resource scheduling, providing key technical support for the high-reliability operation and maintenance of smart grids.

[0042] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method of determining the distance of an obstacle, characterized in that, The method comprises the following steps: acquiring infrared data, depth map data and environment data of a region to be identified; performing identification processing on the infrared data to determine a target obstacle and extract target infrared data of the target obstacle in the infrared data; extracting relevant position information of the target obstacle in the depth map data to obtain three-dimensional point cloud data of the target obstacle; performing analysis processing on the three-dimensional point cloud data to obtain a space main axis direction of the target obstacle, and performing analysis on axial infrared data of the target infrared data along the space main axis direction to obtain a temperature change of the target obstacle; processing the three-dimensional point cloud data to obtain surface curvature of the target obstacle, determining a geometric shape according to the curvature distribution, and determining an initial risk distance according to the geometric shape; correcting the initial risk distance based on the environment data and the temperature change to generate obstacle distance information.

2. The method of determining the distance of an obstacle according to claim 1, wherein, The identification processing on the infrared data to determine a target obstacle and extract target infrared data of the target obstacle in the infrared data comprises the following steps: performing noise reduction preprocessing on the infrared data to obtain smoothed infrared data; segmenting the smoothed infrared data by using an adaptive threshold segmentation technique to obtain a candidate region; performing screening processing on the candidate region based on the area of the candidate region to obtain candidate obstacle region data; inputting the candidate obstacle region data into a preset infrared feature matching algorithm to determine a target obstacle; extracting infrared data in a pixel coordinate range corresponding to the target obstacle in the infrared data to obtain the target infrared data.

3. The method of determining the distance of an obstacle according to claim 1, wherein, The analysis on the axial infrared data of the target infrared data along the space main axis direction to obtain the temperature change of the target obstacle comprises the following steps: extracting a pixel coordinate matrix and a corresponding temperature value matrix from the target infrared data to obtain a structured infrared temperature data set; based on the structured infrared temperature data set and the space main axis direction, converting the space main axis direction into a projection straight line equation in a pixel coordinate system by coordinate transformation to obtain a pixel projection path of the space main axis direction in the target infrared data; extracting temperature data from the target infrared data based on the pixel projection path by using an interpolation algorithm to obtain temperature sequence data; processing the temperature sequence data to obtain the temperature change.

4. The method of determining the distance of an obstacle according to claim 1, wherein, The processing on the three-dimensional point cloud data to obtain the surface curvature of the target obstacle, determining a geometric shape according to the curvature distribution, and determining an initial risk distance according to the geometric shape comprises the following steps: performing noise reduction processing on the three-dimensional point cloud data to obtain denoised three-dimensional point cloud data; selecting a fixed number of neighborhood points for each target point in the denoised three-dimensional point cloud data, fitting a local plane equation of the neighborhood points to obtain local curvature data of each target point; performing analysis and clustering on the local curvature data of each target point to obtain the curvature distribution of the target obstacle; performing analysis on the curvature distribution to determine the geometric shape of the target obstacle; The initial risk distance is obtained by processing the geometric shape according to a preset electric power scene initial risk distance rule base.

5. The method of determining the distance of an obstacle according to claim 1, wherein, The initial risk distance is corrected based on the environment data and the temperature change condition to generate the obstacle distance information of the to-be-identified region. The environment data and the temperature change condition are coupled to determine a correction coefficient of a target obstacle distance. The initial risk distance is corrected based on the correction coefficient to generate the obstacle distance information.

6. A system for determining the distance of an obstacle, characterized in that The method comprises the following steps: An acquisition module is configured to acquire infrared data, depth map data and environment data of a to-be-identified region. An identification module is configured to identify and process the infrared data to determine a target obstacle and extract target infrared data of the target obstacle in the infrared data. An extraction module is configured to extract relevant position information of the target obstacle in the depth map data to obtain three-dimensional point cloud data of the target obstacle. An analysis module is configured to analyze and process the three-dimensional point cloud data to obtain a space main axis direction of the target obstacle, and analyze axial infrared data of the target obstacle along the space main axis direction in the target infrared data to obtain a temperature change condition of the target obstacle. A processing module is configured to process the three-dimensional point cloud data to obtain a surface curvature of the target obstacle, determine a geometric shape according to the curvature distribution, and determine an initial risk distance according to the geometric shape. A generation module is configured to correct the initial risk distance based on the environment data and the temperature change condition to generate obstacle distance information.

7. The system for determining the distance of an obstacle according to claim 6, characterized in that, The identification module comprises: A noise reduction unit is configured to perform noise reduction preprocessing on the infrared data to obtain smoothed infrared data. A segmentation unit is configured to segment the smoothed infrared data by using an adaptive threshold segmentation technique to obtain a candidate region. A screening unit is configured to screen the candidate region based on the area of the candidate region to obtain candidate obstacle region data. A matching algorithm unit is configured to input the candidate obstacle region data into a preset infrared feature matching algorithm to determine a target obstacle. An extraction unit is configured to extract infrared data in a pixel coordinate range corresponding to the target obstacle in the infrared data to obtain the target infrared data.

8. The system for determining the distance of an obstacle according to claim 6, characterized in that, The analysis of the axial infrared data of the target infrared data along the space main axis direction to obtain the temperature change condition of the target obstacle comprises: A structuring unit is configured to extract a pixel coordinate matrix and a corresponding temperature value matrix from the target infrared data to obtain a structured infrared temperature data set. A projection path unit is configured to convert the space main axis direction into a projection straight line equation in a pixel coordinate system by coordinate transformation based on the structured infrared temperature data set and the space main axis direction to obtain a pixel projection path of the space main axis direction in the target infrared data. An extraction unit is configured to extract temperature data from the target infrared data based on the pixel projection path by using an interpolation algorithm to obtain temperature sequence data. A processing unit is configured to process the temperature sequence data to obtain the temperature change condition.

9. The system for determining the distance of an obstacle according to claim 6, characterized in that, The processing module comprises: A denoising unit is configured to perform denoising processing on the three-dimensional point cloud data to obtain denoised three-dimensional point cloud data. A local curvature unit is configured to select a fixed number of neighborhood points for each target point in the denoised three-dimensional point cloud data, fit a local plane equation of the neighborhood points, and obtain local curvature data of each target point. A clustering unit is configured to analyze and cluster the local curvature data of each target point to obtain the curvature distribution of the target obstacle. A curvature unit is configured to analyze the curvature distribution to determine the geometric shape of the target obstacle. A preset unit is configured to process the geometric shape according to a preset initial risk distance rule base of a power scenario to obtain the initial risk distance.

10. The system for determining the distance of an obstacle according to claim 6, characterized in that, The generation module comprises: A coupling unit is configured to perform coupling analysis on the environmental data and the temperature change condition to determine a correction coefficient of a target obstacle distance. A correction unit is configured to correct the initial risk distance based on the correction coefficient to generate the obstacle distance information based on a correction result.

Images