203 results about "Elevation map" patented technology

The invention discloses a hidden disaster-causing factor disaster-causing risk evaluation method based on multi-source data fusion, and relates to the technical field of geographic information systems. The method comprises the following steps of S01, data acquisition, S02, data preprocessing, S03, feature extraction, S04, risk assessment model construction, S05, model training and optimization, S06, risk assessment, S07, result analysis and verification and S08, result output. According to the scheme, by adopting an advanced multi-source data fusion technology, various data sources such as geological data, remote sensing images, elevation maps and hydrological data are effectively integrated; the integration not only widens the coverage range of the data, but also improves the precision and timeliness of the data. Through an automatic data preprocessing process, it is ensured that data of different sources and types can be effectively fused, and a data consistency relation is constructed; the data integration strategy can more comprehensively reflect the complexity and dynamic change of hidden disaster-causing factors, so that the accuracy of disaster-causing risk assessment is remarkably improved.

The invention provides an elevation map path planning method and system based on a terrain adaptive potential energy field, and relates to the technical field of path planning. Comprising the steps of calculating the gradient and roughness of the position of an intelligent agent, setting a terrain adaptive gain, dynamically adjusting the gradient and roughness, and obtaining a terrain adaptive repulsive force field and a composite potential energy field in combination with a basic repulsive force field; determining the motion direction of the agent based on the composite potential energy field, and determining the next position of the agent; in the process that the intelligent agent moves to the next position, whether local re-planning needs to be carried out on the next position of the intelligent agent or not is judged based on a preset environment change judgment rule; and when re-planning is needed, recalculating the terrain adaptive repulsive force field and the dynamic obstacle repulsive force field, updating the composite potential energy field, and re-determining the next position of the intelligent body. According to the invention, the terrain can be effectively sensed, the planning strategy is adaptively adjusted, and high precision, high efficiency and strong robustness are considered.

The invention discloses an image recognition system for industrial part defect detection, and particularly relates to the field of part defect detection, which comprises a multi-axis controllable light source array module, a high-speed polarization camera module, an edge computing node module, a double-branch semantic segmentation network module and a physical constraint post-processing module, according to the invention, through combination of time-sharing stroboscopic illumination and polarization image sequence acquisition, multi-dimensional perception of surface topography and material differences is realized; generating an elevation map and a normal map by using photometric stereo solution, constructing a differential rendering layer reverse matching CAD model, and extracting flash sensitive features; a double-branch U-Net network is adopted to fuse geometric and polarization characteristics, the characterization capability is enhanced through a trans-attention mechanism, and a pixel-level mask is output; and finally, mapping a two-dimensional result to a three-dimensional coordinate system by means of calibration parameters, carrying out geometric verification in combination with a tolerance zone and a height threshold value, and automatically generating a structured defect report containing position, size, grade and visual information.

The invention provides a training method and device for a quadruped robot controller based on reinforcement learning and electronic equipment, and the method comprises the steps: training a first controller according to first sensor data, first elevation map data and a first target control signal, and obtaining a trained first target controller; controlling the quadruped robot to execute a second target control signal in the simulation environment, and determining second sensor data, second elevation map data and RGB image data; constructing training data according to the target position information of each joint, the RGB image data, the second sensor data and the second target control signal; the to-be-trained second controller is trained according to the training data, and the trained second target controller is obtained, so that by using the RGB image data as visual perception information, the method can be more suitable for complex illumination and obstacle environments, and the quadruped robot can be more accurately and efficiently controlled.

The invention discloses a ground element labeling method and device, equipment and a storage medium, and the method comprises the steps: determining time sequence data corresponding to a target scene, and constructing three-dimensional point cloud data of the target scene based on the time sequence data; based on the three-dimensional point cloud data, determining a bird's-eye view and an elevation map corresponding to the three-dimensional point cloud data, and displaying the bird's-eye view on a user interface; in response to the detected labeling operation for the ground elements in the aerial view, determining first labeling information of the ground elements; and determining a labeling result of the ground element based on the first labeling information and the elevation map. According to the technical scheme, the three-dimensional point cloud data of the target scene is constructed, and the aerial view and the elevation map are determined based on the three-dimensional point cloud data, so that semi-automatic manual labeling is performed at the BEV perspective, and meanwhile, the 3D labeling result of the ground element is obtained in combination with the elevation map, so that the high-quality ground element labeling true value can be conveniently and quickly obtained.

The invention provides a mechanical arm real-time stacking model generation method and device based on a 2.5 D elevation map. The method comprises the steps that depth point cloud data of a stacking area is obtained; preprocessing the depth point cloud data to obtain de-noised point cloud data; performing external parameter calibration by adopting Eye-to-hand to obtain an accurate conversion relation between a camera coordinate system and a mechanical arm flange coordinate system, and performing coordinate system conversion to obtain de-noised point cloud data under a unified world coordinate system; performing space clipping on the de-noised point cloud data to obtain effective point cloud data in the stacking area range; a mechanical arm directional bounding box is generated under the unified world coordinate system, and whether the mechanical arm directional bounding box intersects with the stacking area or not is detected through the separation axis theorem; if not, a 2.5 D elevation map is generated according to the effective point cloud data in the stacking area range, and a real-time stacking model of the mechanical arm is obtained according to the 2.5 D elevation map; therefore, quick response and flexible adjustment are realized in a dynamic scene.

The invention discloses a rugged environment improved A * algorithm path planning method based on an elevation map, and relates to the technical field of path planning. The method comprises the following steps: establishing a ground trafficability model and a ground rugged model which conform to the characteristics of a rugged environment terrain based on an elevation map, further identifying the passing difficulty and wave degree of the ground, and then designing an elevation cost function capable of obtaining an optimal path according to the two types of models in combination with the main process of an A * algorithm. And finally, a continuous cost function is established in combination with an original distance cost function, the path optimality is improved, and an optimal route is selected. The improved algorithm enhances the stability and optimality of the planned path of the mobile robot in the rugged environment, and improves the autonomous navigation performance of the mobile robot in the rugged environment.

The invention relates to the field of electric digital data processing, in particular to an automatic modeling method for airport pavement partitioning based on a BIM technology. The invention discloses an automatic modeling method for airport pavement blocks based on a BIM (Building Information Modeling) technology. The automatic modeling method is characterized by comprising the following steps of: 1, processing a block elevation map; 2, decomposing the drawing; 3, creating a self-adaptive family; step 4, picking up; step 5, converting into lines; 6, obtaining a longitudinal starting point; 7, obtaining an ordered transverse line; 8, obtaining a desired point group; step 9, screening and recombination; step 10, obtaining coordinates; eleventhly, coordinates are obtained; step 12, orderly combining; and 13, through nodes, the modeling speed is high.

The invention discloses a quadruped robot landing planning method based on scene decoupling and risk avoidance. The quadruped robot landing planning method aims at solving the problems that in the prior art, perception is not precise in a complex environment, and the obstacle avoidance capacity is insufficient. The method comprises the following steps: acquiring and fusing multi-source sensor data, and segmenting an original point cloud; generating a scene decoupling elevation map of a multi-layer structure by using the segmentation point cloud; constructing a multi-objective optimization problem taking risk avoidance as a core based on the elevation map, wherein the multi-objective optimization problem comprises a comprehensive cost function and a strict obstacle avoidance constraint; utilizing a reaction formula adjusting module of a capturable region theory to cope with a dynamic instability risk; and solving the multi-target and multi-constraint optimization problem in real time by adopting a hierarchical solving strategy, and finally generating an optimal foot end drop point considering safety and stability. According to the method, the terrain adaptability, the motion stability and the decision intelligence of the quadruped robot in an unstructured environment are remarkably improved through fine decoupling of a scene and quantitative avoidance of multi-source risks.

The embodiment of the invention discloses a granary environment three-dimensional modeling method based on three-dimensional point cloud data, and the method comprises the steps: carrying out the non-uniform weighted filtering of the point cloud data in a granary, and obtaining first point cloud data; performing raise dust noise elimination on the first point cloud data based on reflection intensity and dynamic filtering to obtain second point cloud data; identifying the position of the robot through a dynamic shielding detection method, and removing local point cloud data of the position of the robot in the second point cloud data to obtain third point cloud data; classifying hole regions in the third point cloud data, and repairing different types of hole regions; performing down-sampling on the repaired point cloud data by adopting a voxel grid method to generate fourth point cloud data with consistent point density; performing elevation map calculation and flatness analysis on the fourth point cloud data to obtain a granary analysis result; a high-precision granary environment three-dimensional model can be constructed, and environment key information can be extracted.

The invention discloses a method for automatically generating an ancient building elevation by using a multi-modal neural network, and the method comprises the steps: obtaining and preprocessing a three-dimensional inclined point cloud model and a three-dimensional laser point cloud model of a target ancient building, and obtaining the indoor and outdoor point cloud data of the ancient building; performing feature point matching and rigid body transformation fusion on the ancient building indoor and outdoor point cloud data to obtain a three-dimensional ancient building point cloud model; performing semantic segmentation on the three-dimensional ancient building point cloud model through projection and a multi-modal neural network to obtain each structure area of the ancient building; and according to each structural region of the ancient building, extracting each component contour from the three-dimensional point cloud model to generate an ancient building elevation.

The invention discloses a non-flat terrain autonomous navigation method based on path pruning and trajectory smoothing. The method comprises the following steps: constructing an elevation map with three-dimensional map information as a navigation map according to robot navigation space information; based on the navigation map, performing global path search by adopting a path pruning strategy to obtain an initial path from the starting point to the target point; based on the initial path, track optimization is carried out by using a B spline curve, and a smooth track meeting robot dynamics constraints is generated; in response to an obstacle which is detected on the smooth trajectory and cannot be avoided, performing three-dimensional obstacle crossing trajectory planning based on terrain elevation information of the elevation map, and generating an obstacle crossing trajectory capable of crossing the obstacle; wherein the robot performs autonomous navigation based on the smooth track or the obstacle crossing track. According to the invention, efficient global search, smooth trajectory optimization and intelligent obstacle crossing decision of the robot in unknown and non-flat terrains can be realized.

The invention provides a cross-layer path planning method and system based on a multilayer elevation map, and the method comprises the steps: obtaining three-dimensional point cloud data of a multilayer environment, carrying out the adaptive layering of the three-dimensional point cloud data, and obtaining the number of layers of a point cloud group correspondingly covering all elevation information; constructing an elevation map of each layer of point cloud group; calculating the height difference between the current grid and the adjacent grid to obtain the trafficability of the current grid, analyzing the terrain, and optimizing the elevation map; identifying an effective point at the junction of each layer as a passing gateway point; and performing cross-layer path search based on the constructed multi-layer elevation map and the passage gateway point. According to the method, seamless path planning of cross-layer scenes such as stairs and slopes is realized based on the multi-layer elevation map and the layering strategy, and the problem of path breakage during interlayer transition of a traditional method is solved.

The invention relates to the technical field of material transfer, and discloses an unmanned automatic stacking and transferring method and system for bulk materials, and the method comprises the steps: collecting the three-dimensional point cloud data of a bulk material storage yard, carrying out the splicing of the three-dimensional point cloud data, obtaining the overall point cloud of the storage yard, and building an elevation map according to the overall point cloud of the storage yard; collecting body attitude data of the bulk cargo stacking transfer truck, planning a material taking motion path on the basis of the elevation map according to the body attitude data, updating the three-dimensional point cloud data in real time, and correcting the material taking motion path; and communicating with rear-end transfer equipment, performing cooperative control on a material taking motion path, dynamically calculating the quantity of remaining materials in a storage yard according to the three-dimensional point cloud data updated in real time, and completing unmanned automatic stacking, taking and transferring of the bulk materials. Through unmanned and intelligent control, efficient, safe and stable operation of the bucket-wheel stacker-reclaimer is achieved, manual intervention is reduced, the equipment operation efficiency is improved, and the intelligent operation and maintenance requirements of a large bulk material storage yard are met.

An approach is provided for accelerated flood modeling. Using a down-sampling neural network, a low-resolution elevation map is generated from a high-resolution elevation map. Using a partial differential equation (PDE) model, a low-resolution water depth map is generated from the low-resolution elevation map and one or more boundary conditions. Using an up-sampling neural network, a high-resolution water depth map is generated from the low-resolution water depth map. The down-sampling and up-sampling neural networks are trained by minimizing a loss between the generated high-resolution water depth map and ground truth data generated by the PDE model using the high-resolution elevation map as direct input and without using the low-resolution elevation map.

The invention discloses a method and device for determining a road surface unevenness risk area and a storage medium, and belongs to the technical field of road surface evenness detection.The method comprises the steps that initial point cloud data of a detected road surface are acquired; based on a predetermined calibration relation between a point cloud coordinate system and a three-dimensional detection coordinate system, the initial point cloud data is converted into the three-dimensional detection coordinate system to obtain target point cloud data, the transverse axis and the longitudinal axis of the three-dimensional detection coordinate system are parallel to the detected road surface, and the vertical axis of the three-dimensional detection coordinate system is perpendicular to the detected road surface; determining an elevation map of the detected pavement according to the target point cloud data; and determining an uneven risk area of the detected pavement according to the elevation map. According to the detection method based on the point cloud, the point cloud is converted into the elevation map to determine the uneven risky road surface, and the detection method is not affected by environmental factors such as illumination and is high in accuracy.

The invention belongs to the technical field of laser radars. The invention provides a dynamic point cloud detection method and system based on an elevation image. The method comprises the following steps: acquiring ground point cloud, non-ground point cloud and speedometer information of each frame of laser radar point cloud; constructing an elevation image by using the static point cloud and the ground point cloud obtained in the previous frame; carrying out ground continuity restoration on the elevation image; performing shielding detection on the point cloud of the current frame based on an elevation image by combining the odometer information, and determining a dynamic point, a suspected dynamic point and a suspected static point; performing region consistency check on pixels where the dynamic points, the suspected dynamic points and the suspected static points are located, and determining final dynamic points; and performing clustering operation on the detected final dynamic point cloud to obtain a dynamic point cloud detection result. According to the method, the problems of map background point cloud missing, depth image void, easy misjudgment of ground point cloud and the like are effectively solved, so that the accuracy of dynamic object detection is remarkably improved.

A system and method for flood hazard estimation inputs a satellite elevation map and applies a machine learning model to output a geographic map representing flood hazard areas. The machine learning model is trained using a generative adversarial network (GAN) to produce an output of a deterministic hazard mapping algorithm. A GAN objective applies a loss function, reweighted to increase the importance of high hazard areas. The method retrieves a DEM topography file representing elevation data of an identified terrain, and applies a sink-filling algorithm to detect and fill sinks in the DEM topography. The algorithm subtracts the DEM elevation data to generate a filled topography, and identifies flattest regions of the filled topography. The algorithm then generates a flood hazard map by merging the filled topography and the DEM elevation data, using a weighting function that balances the detected sinks and the flattest regions of the filled topography.

The invention provides an unmanned aerial vehicle autonomous homing and obstacle avoidance method, system and device and a medium, and the method comprises the following steps: 1, an unmanned aerial vehicle reads elevation data according to the longitudes and latitudes of a target point and a starting point, and converts the elevation data into three-dimensional coordinates; 2, performing global path exploration by taking elevation data represented by three-dimensional coordinates and unmanned aerial vehicle dynamics constraints as restrictions, and generating a reachable route composed of a plurality of waypoints; step 3, smoothing the reachable route to enable the reachable route to become a better homing route; and 4, homing the unmanned aerial vehicle through the homing route, and avoiding obstacles in the homing process through a laser radar and a DWA algorithm. The method is realized based on an elevation map, and relates to an embedded system, laser radar environment perception and an obstacle avoidance algorithm, so that the unmanned aerial vehicle can safely and quickly realize autonomous homing in an unknown environment.

The invention provides a terrain elevation map generation method and system based on radial basis function fitting, and the method comprises the steps: receiving historical point cloud information and odometer information, generating a local point cloud map, and carrying out the downsampling; dynamically generating a center point of the RBF based on the down-sampled local point cloud map and the ground fitting area; calculating a kernel matrix according to the down-sampled local point cloud map and the central point, and iteratively calculating the weight of the central point through GPU parallel acceleration to generate a terrain manifold; and calculating and outputting an elevation map through GPU parallel acceleration according to the terrain manifold. According to the method, the fitting speed is remarkably increased, the real-time performance of elevation map generation is ensured, and the map generation speed is greatly increased; the elevation information of the dynamic obstacle is updated while the elevation information of the static obstacle is reserved, and perception information is provided for movement, navigation and obstacle avoidance of the intelligent wheelchair; and an elevation map, a gradient map and a normal vector map with any resolution can be output.

According to the hybrid elevation map generation method based on multi-modal data fusion and reinforcement learning, through vision-LiDAR deep fusion, direction perception feature coding, space-time consistency optimization and reinforcement learning dynamic parameter adjustment, elevation map generation with 15-meter long distance and 0.05-meter / pixel high resolution in a medium-speed unmanned driving scene is realized. According to the method, the elevation error is reduced to be lower than 15 cm, the path planning success rate is increased to 90%, the system cost is reduced by 70%, and robust navigation support is provided for a complex unstructured environment.

The invention discloses a point cloud ground segmentation method based on an elevation map, and belongs to the technical field of ground segmentation and modeling. The method comprises the following steps: performing outlier filtering on point cloud data; performing point cloud rasterization on the point cloud data, and calculating a grid point cloud variance; performing ground grid identification on each grid unit according to the grid point cloud data variance; calculating ground grid elevation information to update a local elevation map; non-ground grid elevation information is estimated, and a global elevation map is updated. By adopting the outlier filtering algorithm of the elevation map, the ground points can be dynamically identified according to the surrounding terrain, the false screening of the effective ground points is reduced, and the condition that the point clouds in part of element regions are too sparse is avoided, so that the distribution uniformity and the processing efficiency of the point cloud data are improved, and the requirements on computing resources are reduced; the method is suitable for practical application scenes with limited resources; and a grid point threshold and a variance threshold are dynamically set, so that ground points can be identified and extracted more accurately.

The invention discloses an elevation map and bathymetric map mutual conversion method and system, and the method comprises the steps: obtaining dynamic waterline information and water level information in real time, correcting the dynamic waterline information and the water level information to obtain a corrected instantaneous waterline, and unifying the corrected instantaneous waterline to an elevation point on a target moment waterline; constructing a first Delaunay triangulation network according to the elevation points on the waterline; dividing the first map into an underwater topographic map and a shore topographic map by taking the corrected instantaneous waterline at the target moment as a segmentation line; based on the underwater topographic map, extracting an underwater first point map layer; the first point diagram layer is superposed in the first Delaunay triangulation network; calculating second points corresponding to the first points in the triangular mesh in the first Delaunay triangulation network one by one by adopting an interpolation algorithm; according to the second point, constructing a second Delaunay triangulation network, generating a second line, and further obtaining a corresponding terrain; and superposing the corresponding terrain and the shore topographic map to generate a second map, thereby realizing efficient and accurate conversion.

The invention refers to a computer-implemented method for calibrating a pair of digital elevation maps (DEMI, DEM2), where the digital elevation maps (DEMI, DEM2) of the pair have been acquired from measurements of a first radar sensor equipment (RE1) and indicate for different time points the height of a predetermined terrain (TR) on a surface of a celestial body, wherein the method processes a time series (ts) of SAR images (TSIM) of the predetermined terrain (TR) acquired from measurements of a second radar sensor equipment (RE2), the method comprising the following steps: a) determining candidate calibration points (ccp) from the time series (ts) of SAR images (TSIM) based on dispersion measures (D) of respective pixel positions within the SAR images (TSIM) of the time series (ts); b) identifying pixel positions as calibration points (cp) within the digital elevation maps (DM1, DM2) of the pair based on the candidate calibration points (ccp); c) determining height value differences (Δ) between the digital elevation maps (DEMI, DEM2) of the pair at the calibration points (cp); d) fitting a two-dimensional curve (CS) to the height value differences (A) at the calibration points (cp), the two-dimensional curve (CS) describing the height value differences (Δ) over the pixel positions representing corresponding points within the predetermined terrain (TR); e) correcting the height values (hvl, hv2) of one digital elevation map (DEMI, DEM2) of the pair by the height value differences (Δ) of the two-dimensional curve (CS).

The invention discloses a multi-station time difference and frequency difference fusion positioning method, system and device and a medium, and the method comprises the steps: converting a projection position of a low-orbit observation station on the earth surface into a three-dimensional geographic coordinate in a geodetic coordinate system; and respectively substituting the current radiation source position into the time-frequency difference fusion positioning equation set and the Jacobian matrix of the kth iteration of the time-frequency difference fusion positioning equation set to obtain the residual error of the time-frequency difference fusion positioning equation set and the processed Jacobian matrix, and calculating to obtain the radiation source position after the (k + 1) th iteration, and correcting the height of the radiation source position by using a global elevation map to obtain the radiation source position in the earth-fixed coordinate system, finally replacing the current radiation source position for iteration, and obtaining the optimal estimation value of the radiation source position under the condition that the difference value is smaller than a preset threshold value or the cycle reaches the preset iteration times. And high-precision three-dimensional positioning of a target is realized by combining time difference and frequency difference data and combining a time-frequency difference fusion positioning equation set and global elevation map information.

The invention discloses an unmanned aerial vehicle route global planning method based on key feature points, and relates to the field of unmanned aerial vehicle route planning, and the method comprises the steps: carrying out the hierarchical division of a three-dimensional elevation map according to the performance of an unmanned aerial vehicle; according to obstacle pixel proportions of different layers, an optimal layer is screened out; on the basis of the segmentation maps of the optimal layer and the adjacent layer and in combination with the three-dimensional geographic features, obtaining corrected key feature points; utilizing a heuristic weight calculation formula to calculate heuristic weight values of the corrected key feature points; according to a sorting result of the heuristic weight values, a plurality of corrected key feature points of the unmanned aerial vehicle route are obtained according to a connection barrier-free principle and in combination with a three-dimensional path algorithm until a target point is selected; and based on the initial point, the target point and the plurality of corrected key feature points of the unmanned aerial vehicle route, generating an optimized unmanned aerial vehicle flight path after corner optimization. According to the method, the three-dimensional feasible flight route can be efficiently generated, and the calculation time is short.

The invention discloses a wheat harvesting navigation path generation method and an automatic harvester, and relates to the technical field of automatic driving of agricultural machinery, and the wheat harvesting navigation path generation method comprises the following steps: S1, collecting three-dimensional point cloud data of a wheat field through a laser radar, and enabling a collection region to at least cover the region boundary of a harvested region and a non-harvested region; s2, performing noise reduction and voxel grid downsampling processing on the three-dimensional point cloud data to generate a two-dimensional elevation map; s3, setting grid characteristics and threshold values thereof according to the wheat variety height characteristics; and S4, carrying out binarization processing on the two-dimensional elevation map pixel by pixel based on the grid features and the threshold values thereof. According to the wheat harvesting navigation path generation method, boundary extraction and path conversion are carried out based on radar point cloud, the method has the characteristics of robustness and real-time performance, an automatic harvester based on the method can reduce the recutting rate and land rolling compaction during operation, and resources are effectively saved.

The invention discloses a self-adaptive gesture recognition method based on an FMCW (Frequency Modulated Continuous Wave) radar, aiming at solving the problem that the traditional gesture recognition technology limits the position and orientation of a user. According to the method, an FMCW radar is used for collecting gesture signals, distance, speed, azimuth angle and pitch angle information is extracted through four-dimensional fast Fourier transform, and a distance-Doppler diagram, a distance-azimuth diagram and a distance-pitch angle diagram are constructed. A 2D-CA-CFAR algorithm and a DBSCAN algorithm are adopted to carry out interference suppression and target separation, and parameters of gestures and human body targets are extracted. A parameter mapping algorithm based on spatial physical characteristics is designed, gesture parameters of any position and orientation are mapped to a predefined position and orientation, and position and orientation dependency is eliminated. And finally, calculating a matching distance between each trajectory diagram and the reference trajectory by adopting a fast dynamic time warping algorithm, performing multi-dimensional feature fusion, constructing a fusion matching distance, and constructing a classifier based on the fusion matching distance to realize robust recognition of various gestures in different positions and orientations.

The invention relates to the technical field of image data processing, in particular to a geological safety monitoring and early warning method and system. The method comprises the following steps: acquiring an elevation map of a mining area and a current sailing path of an unmanned aerial vehicle for executing safety exploration in the mining area, and positioning the path on the elevation map; marking pixel points on the path as navigation pixel points, and obtaining a plurality of prediction lines of the navigation pixel points; acquiring predicted elevation data of the navigation pixel points; acquiring corrected elevation data of the navigation pixel points; obtaining collapse coefficients of the navigation pixel points; and according to the collapse coefficient, determining a dangerous area of the mining area, and realizing early warning based on geological safety monitoring. According to the method, prediction is performed based on the navigation pixel points and the historical elevation data on the plurality of prediction lines, so that error transmission caused by single-direction prediction is avoided, the accuracy and reliability of elevation data prediction are improved, and the accuracy of safety monitoring and early warning of a mine is further improved.

The invention provides an underground pipe network detection positioning method and device, and the method comprises the steps: obtaining the multi-source data of an underground pipe network of a to-be-detected point and the surrounding environment, including ground image data, inspection well internal image data, adjacent inspection well spacing data, ground elevation data, pipe bottom elevation data, and position coordinates; inputting the ground image data into a semantic segmentation network, and outputting a ground semantic segmentation image; respectively extracting contours of the ground semantic segmentation image and the image data in the inspection well to obtain a first edge image and a second edge image; constructing a pavement information measurement model; pipeline diameter information, pipeline gradient and rain and sewage pipeline mixed connection conditions are determined; and integrating the pavement information measurement model, the first edge image, the second edge image, the pipeline diameter information, the pipeline gradient and the like, and drawing a planar graph and an elevation graph of the underground drainage pipe network. According to the method, high-precision detection and positioning of the underground pipe network are remarkably improved, and the method is suitable for city updating and city informatization construction.