Search

CN-121982615-A - Mining area road anomaly evaluation method based on pavement data

CN121982615ACN 121982615 ACN121982615 ACN 121982615ACN-121982615-A

Abstract

The application relates to a mining area road anomaly evaluation method based on road surface data, which relates to the technical field of mining safety monitoring and comprises the steps of collecting road video stream in real time and carrying out frame analysis and pretreatment; extracting road surface texture, edge contour and three-dimensional geometric structure characteristics, constructing an evaluation index set containing flatness index, retaining wall integrity coefficient and gap density, identifying retaining wall collapse, road surface gap and structural deformation through a characteristic key point three-dimensional coordinate positioning technology, outputting risk scores by utilizing a pre-training deep learning model, and generating a road integrity evaluation report containing a GIS positioning map and a risk thermodynamic diagram by means of fusion results. The system adopts a collaborative architecture of an edge computing unit and a central server, and integrates a 4K camera set, an FPGA acceleration module and a multi-source data fusion interface. The application breaks through the limitation that the traditional monitoring technology relies on a single-point physical sensor, realizes micro deformation millimeter-level analysis and minute-level response of the road, and remarkably improves the intelligent level of the safety management of the mining area.

Inventors

  • YU WENYANG
  • Si Qianwen

Assignees

  • 北京路凯智行科技有限公司

Dates

Publication Date
20260505
Application Date
20260202

Claims (10)

  1. 1. The mining area road anomaly evaluation method based on the road surface data is characterized by comprising the following steps of: S1, acquiring video stream data of a mining area road in real time, and carrying out frame analysis and pretreatment on the video stream; S2, extracting road feature data in the preprocessed video frame through a computer vision algorithm, wherein the computer vision algorithm adopts a SIFT feature extraction model, and the extracted feature data comprises pavement texture, edge contour and geometric structure information; s3, generating an evaluation index set according to the road characteristic data, wherein the evaluation index set comprises a road surface flatness index, a structural defect density, a road gradient index and a curve passing index; S4, positioning the characteristic key points based on the evaluation index set, and calculating the relative coordinate positions of the characteristic key points in the three-dimensional space; S5, performing multi-category anomaly detection on the characteristic key points, identifying a collapse area of the retaining wall, a pavement gap and a structural deformation part, and recording coordinate data and anomaly grades of the anomaly points; S6, inputting abnormal point location data into a pre-trained deep learning model, wherein the deep learning model adopts a multi-scale convolutional neural network architecture, and outputting a retaining wall collapse probability value and a notch risk score; And S7, fusing the abnormality detection result and the risk score in the step to generate a road integrity assessment report containing an abnormality type locating map, a risk thermodynamic diagram and a maintenance suggestion.
  2. 2. The mining area road anomaly evaluation method based on road surface data according to claim 1, wherein the preprocessing in step S1 comprises defogging and enhancing the video stream, compensating image degradation in a mining area dust environment by adopting a dark channel prior algorithm, and eliminating dynamic shielding interference caused by transport vehicles through time-space domain filtering.
  3. 3. The mining area road abnormality assessment method based on road surface data according to claim 1, wherein the calculation of the structural defect density in step S3 includes: Obtaining a three-dimensional coordinate set of a characteristic point group and comparing the three-dimensional coordinate set with a reference model, calculating the space position offset of each point, normalizing and calculating an arithmetic mean value to obtain a structural defect coefficient, wherein the coefficient range is 0-1; Counting the gap density, namely counting the number of abnormal points which accord with the gap size threshold value in a unit area; Structural defect density = structural defect coefficient x 0.7+ notch density x 0.3.
  4. 4. The mining area road abnormality assessment method based on road surface data according to claim 1, wherein the determination of the relative coordinate position in step S4 includes establishing a local coordinate system based on a road center line, calculating three-dimensional coordinates of feature key points by stereoscopic vision matching, and labeling the height deviation value of each key point with respect to a retaining wall base plane.
  5. 5. The mining area road abnormality assessment method based on road surface data according to claim 1, wherein the multi-category abnormality detection in step S5 specifically includes: The collapse detection of the retaining wall, namely judging a collapse area when the height deviation of more than 3 continuous characteristic key points exceeds a 15cm threshold value; detecting a notch, namely identifying an edge contour discontinuous region and triggering an alarm when the width of the notch exceeds 30cm and the depth exceeds 20 cm; structural deformation detection, namely analyzing outline curvature abnormal change through Fourier descriptor.
  6. 6. The mining area road anomaly evaluation method based on road surface data according to claim 1, wherein the training of the deep learning model in step S6 uses transfer learning to initialize ResNet-101 backbone network parameters, fine-tunes using a dataset containing 10 ten thousand sets of mining area road anomaly samples, and applies a focus loss function to solve the sample imbalance problem.
  7. 7. The mining area road abnormality evaluation method based on road surface data according to claim 1, wherein the road integrity evaluation report in step S7 includes: an abnormal position geographic information system label graph; a three-dimensional road model partitioned according to the risk level; And predicting curves based on the abnormal development trend of the historical data.
  8. 8. An intelligent mining area roadway monitoring system for implementing the method of any one of claims 1-7, comprising: a vehicle-mounted 4K ultra-high definition camera set (101) mounted on a mining area inspection vehicle, wherein the camera set is provided with an infrared light supplementing module; the edge computing unit (102) is internally provided with a GPU acceleration card and is used for processing video stream data in real time; a central analysis server (103) running the deep learning model and an assessment report generation module; and the mobile terminal alarm device (104) is used for pushing the emergency treatment instruction.
  9. 9. The mining area road intelligent monitoring system according to claim 8, characterized in that the edge calculation unit (102) is integrated with a video decoding chip supporting h.265 hard decoding, a special FPGA acceleration module for road feature extraction and a secure storage area for locally caching anomaly data.
  10. 10. The mining area road intelligent monitoring system according to claim 8, wherein the central analysis server (103) is configured with a reference database storing standard road three-dimensional models under different geological conditions, an adaptive calibration module for periodically updating feature key point reference coordinates, and a multi-source data fusion interface for accessing transport vehicle load sensor data.

Description

Mining area road anomaly evaluation method based on pavement data Technical Field The application relates to the technical field of mining safety monitoring, in particular to a mining area road abnormality assessment method and system based on computer vision and deep learning, which are suitable for real-time monitoring of structural abnormalities such as subsidence, gaps and the like of mining area roads such as strip mines, quarries and the like. Background Along with the acceleration of the automation process of mining industry, mining area road safety has become a core link for guaranteeing mining efficiency and personnel safety, and the current mainstream monitoring means mainly depend on two technologies, namely a video monitoring system for collecting road images through a fixed camera and a manual inspection mechanism for relying on a safety officer to periodically inspect on site. At present, china patent application number CN202410887430.3 discloses a mining area earth surface coverage change and subsidence area monitoring system, which comprises a rainfall monitoring module, a GPS sedimentation monitoring module, a communication module and a monitoring center, wherein the rainfall monitoring module is used for collecting mining area rainfall information, the GPS sedimentation monitoring module is used for collecting crack body surface displacement information, the communication module is used for sending the collected rainfall information and the crack body surface displacement information to the monitoring center, the monitoring center is used for storing and analyzing data and automatically displaying graphics, the mining area rainfall information comprises rainfall rate, hour rainfall and daily rainfall, and the crack body surface displacement information comprises daily average displacement rate, displacement direction and accumulated displacement. However, the prior art mainly relies on a monitoring mode of single-point layout of a physical sensor, is difficult to capture three-dimensional microscopic deformation characteristics of road structure anomalies such as collapse of a retaining wall and pavement gaps, particularly in complex scenes such as dynamic shielding of vehicles in mining areas and dislocation of underground strata, GPS displacement data can only reflect accumulated displacement of the ground surface and cannot analyze the structural integrity degradation process, meanwhile, data analysis delay exists in passive monitoring of surface displacement of a crack body by a system, sudden road collapse risks are not conveniently identified in real time, and a linkage evaluation mechanism for the safety state of roads around the collapse areas is also lacked, so that safety early warning and road maintenance decision are split. Disclosure of Invention The application aims to provide a mining area road abnormality assessment method based on road surface data, so as to solve the problems in the background technology. In a first aspect, the mining area road abnormality assessment method based on road surface data provided by the application adopts the following technical scheme that the mining area road abnormality assessment method comprises the following steps: S1, acquiring video stream data of a mining area road in real time, and carrying out frame analysis and pretreatment on the video stream; S2, extracting road feature data in the preprocessed video frame through a computer vision algorithm, wherein the computer vision algorithm adopts a SIFT feature extraction model, and the extracted feature data comprises pavement texture, edge contour and geometric structure information; s3, generating an evaluation index set according to the road characteristic data, wherein the evaluation index set comprises a road surface flatness index, a structural defect density, a road gradient index and a curve passing index; S4, positioning the characteristic key points based on the evaluation index set, and calculating the relative coordinate positions of the characteristic key points in the three-dimensional space; S5, performing multi-category anomaly detection on the characteristic key points, identifying a collapse area of the retaining wall, a pavement gap and a structural deformation part, and recording coordinate data and anomaly grades of the anomaly points; S6, inputting abnormal point location data into a pre-trained deep learning model, wherein the deep learning model adopts a multi-scale convolutional neural network architecture, and outputting a retaining wall collapse probability value and a notch risk score; And S7, fusing the abnormality detection result and the risk score in the step to generate a road integrity assessment report containing an abnormality type locating map, a risk thermodynamic diagram and a maintenance suggestion. Preferably, the preprocessing in step S1 includes defogging and enhancing the video stream, compensating image degradation in the dust-raising environm