A method for detecting open-pit mine disasters based on drones

Abstract

The present invention discloses an open-pit mine disaster detection method based on a drone, which belongs to the field of mine disaster detection technology, including: evaluating terrain complexity and geological stability, obtaining known disaster risk area information, and generating a risk area distribution map based on the known disaster risk area information; obtaining high-risk areas based on terrain complexity, geological stability and risk area distribution map, and generating a drone flight path that avoids high-risk areas based on the high-risk areas; collecting high-resolution image data in real time through a drone according to the drone flight path to obtain a panoramic image map; using a deep learning algorithm to detect the panoramic image map to obtain attribute information of a suspected collapse area; combining known mine collapse feature data, using a similarity matching algorithm to judge the authenticity of the collapse area, and obtaining a detection result; uploading the detection result to a management platform for real-time display and dynamic update. The present invention improves the efficiency and accuracy of mine disaster monitoring.

Description

Technical Field

[0001] The present invention belongs to the technical field of mine disaster detection, and in particular relates to an open-pit mine disaster detection method based on an unmanned aerial vehicle. Background Art

[0002] When drones conduct disaster detection in open-pit mines, the drone disaster avoidance path planning technology can independently plan flight paths through comprehensive analysis of mine terrain, geological conditions, disaster risk areas and other information to avoid high-risk areas and ensure the flight safety of drones and the effectiveness of disaster detection. The collapse information technology focuses on the information collection, analysis and management of mine collapse areas, providing important data support for drone disaster detection, but lacks consideration in the specific drone flight path planning. The two can be combined to achieve more comprehensive and accurate mine disaster detection. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes to solve the problems existing in the above prior art.

[0004] To achieve the above object, the present invention provides an open-pit mine disaster detection method based on a drone, comprising the following steps:

[0005] Obtaining topographic and geological structure parameters of the mining area, assessing terrain complexity and geological stability, obtaining information on known disaster risk areas, and generating a risk area distribution map based on the known disaster risk area information;

[0006] Obtaining high-risk areas based on the terrain complexity, geological stability and risk area distribution map, and generating a UAV flight path that avoids the high-risk areas based on the high-risk areas;

[0007] Collect high-resolution image data in real time using the drone according to the drone flight path to obtain a panoramic image;

[0008] Using a deep learning algorithm to detect the panoramic image to obtain attribute information of suspected collapsed areas;

[0009] Combined with known mine collapse feature data, a similarity matching algorithm is used to determine the authenticity of the collapse area and obtain the detection results;

[0010] Upload the test results to the management platform for real-time display and dynamic updates.

[0011] Preferably, the method for evaluating terrain complexity and geological stability based on the topography and geological structure parameters of the mining area includes:

[0012] Using geographic information system technology, according to the topographical features and geological structure parameters, the terrain complexity of the target mining area is quantitatively analyzed to obtain terrain complexity analysis results;

[0013] A stability assessment model is used to quantitatively assess the geological stability of the target mining area based on the geological structure parameters to obtain a geological stability assessment result.

[0014] Preferably, the method for generating a risk area distribution map comprises:

[0015] The known disaster risk area information is obtained, the disaster risk area information is structured and converted into a unified data format; based on the location coordinate data of the disaster risk area, it is determined whether the disaster risk area is located within the boundary of the target mining area, and the disaster risk area data located in the target mining area are screened out; the disaster risk area data is preprocessed, the risk level and disaster type attributes are extracted and converted into numerical feature vectors; the K-means clustering algorithm is used to perform cluster analysis on the numerical feature vectors, and the disaster risk area is divided into different risk levels according to the clustering results to generate a disaster area distribution image.

[0016] Preferably, the method for generating the UAV flight path comprises:

[0017] Based on the terrain complexity, geological stability and risk area distribution map, a UAV flight environment model is constructed; an A heuristic search algorithm is used to search for UAV flight paths in the flight environment model to generate an initial flight route; and the initial flight route is optimized to avoid high-risk areas according to the risk area distribution map to obtain an optimized flight path.

[0018] Preferably, the method for obtaining a panoramic image includes:

[0019] The flight trajectory data and attitude data of the UAV are obtained, and the spatial position and attitude angle of the UAV are determined based on the flight trajectory data and attitude data; the collected original image data are preprocessed to obtain a preprocessed image; according to the spatial position and attitude angle of the UAV, the overlapping area between adjacent images in the preprocessed image is calculated, and the feature point matching algorithm is used to extract the feature points in the overlapping area, and the correspondence between the feature points is calculated; according to the correspondence between the feature points, a random sampling consistency algorithm is used to estimate the homography matrix between adjacent images, and the homography matrix is ​​used to perform geometric correction and registration on the images; the registered image data is input into the image stitching model, and a multi-band fusion algorithm is used to fuse the image data in the frequency domain to eliminate the brightness difference and color difference between the images, and obtain a seamlessly stitched panoramic image.

[0020] Preferably, when collecting raw image data, the exposure parameters of the airborne optical sensor are controlled to obtain the raw image data collected by the sensor, and the exposure time and gain coefficient of the sensor are dynamically adjusted according to the brightness distribution of the image data.

[0021] Preferably, the preprocessing method comprises: using a radial distortion correction model to eliminate radial distortion and tangential distortion of the image data according to intrinsic parameters and extrinsic parameters of the optical sensor.

[0022] Preferably, the method for obtaining attribute information of a suspected collapse area comprises:

[0023] The image is detected in real time through the pre-trained YOLOv5 target detection model. When a suspected collapse area is detected, the suspected collapse area is located through a deep learning algorithm to obtain the coordinate information of the suspected collapse area, and the geographical location is obtained based on the coordinate information. The edge of the located suspected collapse area is extracted through an image segmentation algorithm to obtain the contour information of the suspected collapse area; based on the contour information, the area size of the suspected collapse area is calculated using the polygon area calculation formula; the coordinate information, geographical location and area size are associated and stored to form structured collapse area attribute information.

[0024] Compared with the prior art, the present invention has the following advantages and technical effects:

[0025] The present invention discloses a method for detecting open-pit mine disasters based on unmanned aerial vehicles. First, by obtaining the topographic and geomorphic characteristic parameters and geological structure characteristic parameters of the target mining area, the terrain complexity and geological stability evaluation results of the mining area are determined by comprehensive analysis. Then, the known disaster risk areas are intelligently classified by using a clustering algorithm to generate a distribution map of disaster areas with different risk levels. Based on these evaluation results and distribution maps, a heuristic search algorithm is used to automatically generate the flight path of the unmanned aerial vehicle, and the path is optimized to avoid high-risk areas to ensure flight safety and full coverage of disaster detection. During the flight, the airborne optical sensor collects high-resolution image data in real time, and obtains a panoramic image of the mining area through an image stitching algorithm. Subsequently, a target detection algorithm based on deep learning detects and locates suspected collapse areas in real time, and combines the mine collapse feature database to accurately identify the collapse area through a similarity matching algorithm. Finally, the detected collapse area information is uploaded to the mine disaster information management platform to realize the real-time display and dynamic update of the collapse area, providing data support for mine disaster warning and emergency decision-making. The present invention improves the efficiency and accuracy of mine disaster monitoring through intelligent path planning, real-time image stitching and collapse detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0027] Figure 1 The figure is a flow chart of a method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0029] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0030] Embodiment 1

[0031] like Figure 1 As shown, this embodiment provides an open-pit mine disaster detection method based on a drone, comprising the following steps:

[0032] Obtain the topographic and geological structure parameters of the mining area, evaluate the terrain complexity and geological stability, obtain information on known disaster risk areas, and generate a risk area distribution map based on the known disaster risk area information;

[0033] Based on the terrain complexity, geological stability and risk area distribution map, high-risk areas are obtained, and based on the high-risk areas, UAV flight paths that avoid high-risk areas are generated;

[0034] Collect high-resolution image data in real time through the drone according to the flight path of the drone to obtain a panoramic image map;

[0035] Use deep learning algorithms to detect panoramic images and obtain attribute information of suspected collapse areas;

[0036] Combined with known mine collapse feature data, a similarity matching algorithm is used to determine the authenticity of the collapse area and obtain the detection results;

[0037] Upload the test results to the management platform for real-time display and dynamic updates.

[0038] The specific implementation methods are as follows:

[0039] S101. Based on the pre-established database of mining terrain and geological conditions, obtain the terrain and geomorphic characteristic parameters and geological structure characteristic parameters of the target mining area, and determine the terrain complexity and geological stability assessment results of the mining area through comprehensive analysis.

[0040] Obtain topographic and geomorphic parameters and geological structure parameters of the target mining area;

[0041] Using geographic information system technology, according to the topographic parameters and geological structure parameters, the terrain complexity of the target mining area is quantitatively analyzed to obtain terrain complexity analysis results;

[0042] Using a stability assessment model, based on the geological structure parameters, a quantitative assessment of the geological stability of the target mining area is performed to obtain a geological stability assessment result;

[0043] By using data fusion technology, the terrain complexity analysis results and the geological stability assessment results are integrated to obtain a comprehensive terrain and geological stability assessment report of the target mining area;

[0044] Specifically, firstly, the topographic parameters such as elevation, slope, and slope direction of the target mining area, as well as geological structural parameters such as lithology, strata, and faults are obtained through drone aerial photography and satellite remote sensing. Then, ArcGIS software is used to perform spatial interpolation analysis on the acquired data to generate a digital elevation model of the mining area, and to calculate indicators such as terrain undulation and elevation difference, quantitatively evaluate the complexity of the mining area terrain, and obtain the analysis results of the terrain complexity. At the same time, the fuzzy comprehensive evaluation method is used to score the geological structural stability of the mining area from the aspects of lithology strength, stratum occurrence, fault density, etc., and the geological stability assessment value is calculated. Then, the weighted average method is used to assign weight coefficients to the two results of terrain complexity and geological stability, and the comprehensive terrain and geological stability assessment values ​​of the mining area are obtained by comprehensive calculation.

[0045] S102. For the target mining area, obtain the location coordinates, risk level, disaster type and other attribute information of the pre-marked known disaster risk areas, perform intelligent classification of the disaster risk area information through a clustering algorithm, and obtain a disaster area distribution map of different risk levels.

[0046] The geographical location information and boundary range data of the target mining area are obtained, and converted into a polygonal area in a spatial coordinate system; the location coordinates, risk level, and disaster type attribute information of the pre-marked known disaster risk areas are obtained, the attribute information is structured and converted into a unified data format; according to the location coordinate data of the disaster risk area, it is determined whether each of the disaster risk areas is located within the boundary range of the target mining area, and the disaster risk area data located in the target mining area are screened out; the screened disaster risk area data are preprocessed, the risk level and the disaster type attributes are extracted, and the data are converted into a numerical feature vector; the K-means clustering algorithm is used to perform cluster analysis on the disaster risk area feature vectors, and the disaster risk areas are divided into different risk levels according to the clustering results; the clustered disaster risk area data are spatially visualized, and disaster areas of different risk levels are marked with different colors or patterns on the geographic base map of the target mining area; a disaster area distribution image is generated according to the visualization result, and the image is associated with the attribute information of the disaster risk area to form a complete disaster risk area distribution map data set.

[0047] Specifically, first, the geographical coordinates of the target mining area are obtained by using GPS positioning technology, such as longitude 1123°E and latitude 354°N. Then, the mining area boundary is extracted by remote sensing image interpretation technology to obtain a polygonal area composed of multiple coordinate points. Then, the location coordinates, hazard level, and disaster type of known disaster points such as landslides and debris flows in the area are obtained from the existing disaster information database, and converted into a unified CSV format through Python's Pandas library. Using the spatial query function of ArcGIS, the mining area boundary polygon is used as the query condition to filter out the disaster point data located inside the mining area. These data are preprocessed to extract the two attributes of hazard level (such as high risk, medium risk, and low risk) and disaster type (such as landslide and debris flow), which are assigned to 1, 2, 3 and 0, 1 respectively, and converted into numerical feature vectors. Then, the K-means clustering algorithm is used to set the cluster number k=3, and the feature vector is clustered to obtain 3 cluster centers, representing the three risk levels of high risk, medium risk, and low risk, respectively. Then, using ArcGIS's symbolization function, we used red, orange, and yellow to represent high, medium, and low risk areas, respectively, to generate a disaster risk distribution map. Finally, we spatially connected the distribution map with the attribute information table of the disaster point to generate a vector dataset containing information such as location, type, and hazard level, providing data support for subsequent disaster prevention and control decisions.

[0048] The K-means clustering algorithm is used to perform cluster analysis on the characteristic vectors of the disaster risk area, and the disaster risk area is divided into different risk levels according to the clustering results.

[0049] Obtain geographic information data and historical disaster data of disaster risk areas, extract characteristic parameters of disaster risk areas, and construct characteristic vectors of disaster risk areas;

[0050] The characteristic vector of the disaster risk area is normalized by using data preprocessing technology to eliminate the influence of the difference in dimension and value range of different characteristic parameters;

[0051] According to the dimension and data distribution characteristics of the characteristic vector of the disaster risk area, the number of cluster centers and the threshold of the number of iterations of the K-means clustering algorithm are set;

[0052] Applying the K-means clustering algorithm to perform cluster analysis on the preprocessed disaster risk area feature vectors to obtain clustering results of the disaster risk areas;

[0053] Based on the K-means clustering results, the average risk level of the disaster risk area feature vector contained in each cluster center is calculated to obtain the risk level of each cluster;

[0054] According to the risk level of each cluster, the clustering results are sorted to determine the risk level represented by different clusters;

[0055] The disaster risk area is divided into corresponding risk levels according to the cluster to which its characteristic vector belongs, and a risk level division map of the disaster risk area is generated.

[0056] S103. Based on the complexity of the mine terrain, the geological stability assessment results and the disaster risk area distribution map, a heuristic search algorithm is used to automatically generate the flight path of the UAV, and the path is optimized to avoid high-risk areas, ensuring the flight safety of the UAV and full coverage of disaster detection.

[0057] According to the terrain complexity index, geological stability level and risk area distribution map, a UAV flight environment model is constructed; the A heuristic search algorithm is used to search the UAV flight path in the flight environment model and generate an initial flight route; the initial flight route is optimized and the path is adjusted according to the risk area distribution map to avoid high-risk areas and obtain an optimized flight route; the optimized flight route is sent to the UAV, and the UAV is controlled to fly along the route to achieve full coverage detection of mine disasters.

[0058] Specifically, the terrain complexity index, geological stability level and risk area distribution layer are rasterized to construct a three-dimensional space grid model as the UAV flight environment model. On this basis, the UAV take-off and landing points are used as the starting and end points, and the flight altitude and speed are used as constraints. The A* heuristic search algorithm is used to search the UAV flight path to obtain the initial route. Then, the path nodes in the high-risk area are identified by comparing with the risk area distribution map. The risk avoidance factor is introduced through the artificial potential field method to make the flight route deviate from the high-risk area, and the smoothness and accessibility of the path are guaranteed, so as to obtain the optimized flight route. Finally, the route coordinate points are converted into waypoint format and sent to the UAV flight control system, which flies autonomously and obtains the image data of mine disasters to achieve full coverage and no dead angle detection.

[0059] S104. During the flight of the UAV, high-resolution image data is collected in real time by using the airborne optical sensor, and the collected images are stitched in real time using an image stitching algorithm to obtain a panoramic image of the mining area.

[0060] Obtain the flight trajectory data and attitude data of the drone, and determine the spatial position and attitude angle of the drone based on the flight trajectory data and attitude data; control the exposure parameters of the onboard optical sensor, obtain the original image data collected by the sensor, and dynamically adjust the exposure time and gain coefficient of the sensor based on the brightness distribution of the image data; perform distortion correction processing on the collected original image data, and use the radial distortion correction model based on the intrinsic and extrinsic parameters of the sensor to eliminate the radial distortion and tangential distortion in the image data; calculate the overlapping area between adjacent images based on the spatial position and attitude angle of the drone, and use the feature point matching algorithm to extract the overlapping area The feature points in the image are identified and the corresponding relationship between the feature points is calculated. According to the corresponding relationship between the feature points, the random sampling consistency algorithm is used to estimate the homography matrix between adjacent images, and the homography matrix is ​​used to perform geometric correction and registration on the images. The registered image data is input into the image stitching model, and the multi-band fusion algorithm is used to fuse the image data in the frequency domain to eliminate the brightness difference and color difference between the images and obtain a seamlessly stitched panoramic image. The stitched panoramic image is orthorectified and the collinear equation model is used to convert the panoramic image into an orthophoto image according to the altitude data and camera parameters of the drone to eliminate the geometric deformation of the image.

[0061] When the UAV performs the task of mine disaster detection, the flight trajectory data and attitude data of the UAV are first obtained through GPS and gyroscope to ensure that the spatial position and attitude angle of the UAV can be accurately calculated. For example, the absolute coordinates of the UAV in the air are (350m, 400m, 50m), and the attitude angles are 30 degrees yaw, 15 degrees pitch and 5 degrees roll. Subsequently, the exposure parameters of the onboard optical sensor are controlled, and the exposure time of the sensor is dynamically adjusted to 1 / 200 seconds and the gain coefficient is 5dB according to the brightness distribution of the real-time collected image data to adapt to different lighting conditions. The collected raw image data is subjected to distortion correction processing, and a radial distortion correction model is used, such as the cv::undistort function in the OpenCV library, to eliminate radial distortion and tangential distortion in the image data and improve the image quality. The spatial position and attitude angle of the UAV are used to calculate the overlapping area between adjacent images, and the feature points in the overlapping area are extracted by the SIFT algorithm, and the corresponding relationship between the feature points is calculated. The random sampling consensus algorithm (RANSAC) is used to estimate the homography matrix between adjacent images, and then the matrix is ​​used to perform geometric correction and registration on the images. The registered image data is input into the image stitching model, and the multi-band fusion algorithm is used to fuse the image data in the frequency domain to eliminate the brightness difference and color difference between the images. For example, the Laplacian pyramid is used for multi-scale fusion to ensure seamless stitching of panoramic images. Finally, the stitched panoramic images are orthorectified. According to the altitude data and camera parameters of the drone, the collinear equation model is used to convert the panoramic images into orthophoto images to eliminate the geometric deformation of the images and ensure that the geographical location of the images accurately corresponds to the actual terrain.

[0062] S105. For the stitched panoramic image of the mining area, a target detection algorithm based on deep learning is used to detect and locate the suspected collapse area in real time, and obtain the corresponding attribute information such as location coordinates and area.

[0063] According to the pre-established deep learning target detection model, the acquired panoramic image of the mining area is detected in real time to determine whether there is a suspected collapse area; if a suspected collapse area is detected, the area is accurately located through the deep learning algorithm to obtain its position coordinate information; based on the obtained position coordinate information, combined with the preset mining area geographic information system, the specific geographical location of the suspected collapse area is determined; through the image segmentation algorithm, the edge of the located suspected collapse area is extracted to obtain the contour information of the area; based on the extracted contour information, the polygon area calculation formula is used to calculate the area size of the suspected collapse area; the attribute information such as the location coordinates, geographical location and area size of the suspected collapse area are associated and stored to form structured collapse area attribute data.

[0064] Specifically, after obtaining the panoramic image of the mining area, the pre-trained YOLOv5 target detection model is used to detect the image in real time. The model uses DarkNet-53 as the backbone network, is pre-trained on the COCO dataset, and is fine-tuned on a self-built dataset containing 10,000 images of mining area collapse areas, and finally achieves 95% mAP on the validation set. Through this model, it is possible to determine in real time whether there is a suspected collapse area in the image. If a suspected collapse area is detected, the Mask R-CNN algorithm is used to accurately locate and semantically segment the area to obtain its pixel-level location coordinate information and edge contour information. The Mask R-CNN algorithm adds the FPN structure and RoIAlign alignment method on the basis of ResNet-101, improves the accuracy of feature extraction and regional positioning, and achieves 92% mIoU on the self-built mining area collapse area segmentation dataset. According to the acquired location coordinate information, combined with the preset mining area geographic information system, the specific geographical location of the suspected collapse area can be determined through coordinate conversion and spatial mapping, and the positioning accuracy can reach the centimeter level. According to the extracted edge contour information, the polygon area calculation formula A=5*[(x1*y2-x2*y1)+(x2*y3-x3*y2)+...+(xn*y1-x1*yn)] can be used to calculate the exact area size of the suspected collapse area. Finally, the location coordinates, geographical location, area size and other attribute information of the suspected collapse area are associated and stored to form structured collapse area attribute data.

[0065] S106. Based on the detected suspected subsidence area information and in combination with a pre-established mine subsidence feature database, a similarity matching algorithm is used to determine the authenticity of the suspected subsidence area, and an accurate identification result of the subsidence area is obtained.

[0066] Based on the detected suspected collapse area information, extract the characteristic data of the area, including collapse shape, area, depth and other parameters; compare the extracted characteristic data with the data in the pre-established mine collapse feature database to calculate the similarity; use the similarity matching algorithm to judge the authenticity of the suspected collapse area based on the calculated similarity value; if the similarity exceeds the preset threshold, the suspected collapse area is determined to be a real collapse area, otherwise it is determined to be a false alarm; for the area determined to be a real collapse area, further obtain its precise boundary range and collapse degree and other information; output the precise identification results, and update the mine collapse feature database to optimize the accuracy of subsequent identification; based on the identification results, use machine learning algorithms such as support vector machines or random forests to automatically classify and risk assess the mine collapse area.

[0067] Specifically, in actual operation, the deep learning model is first used to extract features of suspected collapse areas in the panoramic image of the mining area, including the shape, area (in square meters) and depth (in meters) of the collapse. These feature data are compared with the historical data stored in the mine collapse feature database, and the cosine similarity algorithm is used to calculate the similarity between the two. If the calculated similarity value exceeds the preset threshold (such as 8), the suspected collapse area is considered to be a real collapse area. For the confirmed real collapse area, the image segmentation technology is further used to accurately obtain its boundary range, and the degree of collapse is determined by geological analysis. This result will be output and used to update the mine collapse feature database to improve the accuracy of future identification. Finally, machine learning algorithms such as support vector machines or random forests are used to classify and assess the risks of the collapse area. These algorithms can conduct a comprehensive analysis based on multiple features such as the area, depth and shape of the collapse, and output the risk level of each area, providing a scientific basis for safety management and disaster prevention in the mining area.

[0068] Machine learning algorithms such as support vector machines or random forests are used to automatically classify and assess the risks of mine subsidence areas.

[0069] S107. Upload the location coordinates, area, identification confidence and other information of the subsidence area detected by the drone to the mine disaster information management platform, and realize real-time display and dynamic update of the subsidence area information through a visual interface, so as to provide data support for mine disaster warning and emergency decision-making.

[0070] The visualization module of the mine disaster information management platform reads the information of the subsidence area and displays the location and range of the subsidence area on the electronic map in real time through WebGIS and other technologies. The data update module of the mine disaster information management platform regularly obtains the latest subsidence area information from drones and dynamically updates the subsidence area information on the platform. The mine disaster information management platform determines whether disaster warning is needed based on the location, area, recognition confidence and other information of the subsidence area, combined with the preset warning threshold, and provides data support for mine disaster emergency decision-making.

[0071] Specifically, the platform uses a distributed database to store and manage massive amounts of drone data, and uses message queues such as Kafka for decoupling and asynchronous processing to ensure high availability and real-time performance of the system. Through WebGIS frameworks such as Leaflet and Cesium, the location and range information of the subsidence area is dynamically rendered on the electronic map, and interactive operation functions are provided to facilitate user query and analysis. The data update module sets a scheduled task to obtain the latest subsidence area information from the drone every 30 minutes, and synchronizes the data on the platform through incremental updates to ensure that the displayed information is consistent with the actual situation. According to parameters such as the area of ​​the subsidence area and the recognition confidence, different levels of warning thresholds are set. When the subsidence area of ​​a certain area exceeds 1,000 square meters or the confidence is higher than 95%, the warning information is pushed in a timely manner, and corresponding emergency response suggestions are given in combination with the expert knowledge base, providing strong support for the prevention and control of mine disasters and decision-making.

[0072] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for detecting open-pit mine disasters based on drones, characterized in that: The following steps are involved: Obtaining topographic and geological structure parameters of the mining area, assessing terrain complexity and geological stability, obtaining information on known disaster risk areas, and generating a risk area distribution map based on the known disaster risk area information; Obtaining high-risk areas based on the terrain complexity, geological stability and risk area distribution map, and generating a UAV flight path that avoids the high-risk areas based on the high-risk areas; Collect high-resolution image data in real time using the drone according to the drone flight path to obtain a panoramic image; Using a deep learning algorithm to detect the panoramic image to obtain attribute information of suspected collapsed areas; Combined with known mine collapse feature data, a similarity matching algorithm is used to determine the authenticity of the collapse area and obtain the detection results; Upload the test results to the management platform for real-time display and dynamic updates.

2. The open-pit mine disaster detection method based on drone according to claim 1 is characterized in that: Methods for assessing terrain complexity and geological stability include: Using geographic information system technology, according to the topographical features and geological structure parameters, the terrain complexity of the target mining area is quantitatively analyzed to obtain terrain complexity analysis results; A stability assessment model is used to quantitatively assess the geological stability of the target mining area based on the geological structure parameters to obtain a geological stability assessment result.

3. The open-pit mine disaster detection method based on drone according to claim 1 is characterized in that: Methods for generating risk area distribution maps include: The known disaster risk area information is obtained, the disaster risk area information is structured and converted into a unified data format; based on the location coordinate data of the disaster risk area, it is determined whether the disaster risk area is located within the boundary of the target mining area, and the disaster risk area data located in the target mining area are screened out; the disaster risk area data is preprocessed, the risk level and disaster type attributes are extracted and converted into numerical feature vectors; the K-means clustering algorithm is used to perform cluster analysis on the numerical feature vectors, and the disaster risk area is divided into different risk levels according to the clustering results to generate a disaster area distribution image.

4. The open-pit mine disaster detection method based on drone according to claim 1 is characterized in that: The method for generating the flight path of the drone includes: Based on the terrain complexity, geological stability and risk area distribution map, a UAV flight environment model is constructed; an A heuristic search algorithm is used to search for UAV flight paths in the flight environment model to generate an initial flight route; and the initial flight route is optimized to avoid high-risk areas according to the risk area distribution map to obtain an optimized flight path.

5. The open-pit mine disaster detection method based on drone according to claim 1 is characterized in that: The method for obtaining a panoramic image comprises: The flight trajectory data and attitude data of the UAV are obtained, and the spatial position and attitude angle of the UAV are determined based on the flight trajectory data and attitude data; the collected original image data are preprocessed to obtain a preprocessed image; according to the spatial position and attitude angle of the UAV, the overlapping area between adjacent images in the preprocessed image is calculated, and the feature point matching algorithm is used to extract the feature points in the overlapping area, and the correspondence between the feature points is calculated; according to the correspondence between the feature points, a random sampling consistency algorithm is used to estimate the homography matrix between adjacent images, and the homography matrix is ​​used to perform geometric correction and registration on the images; the registered image data is input into the image stitching model, and a multi-band fusion algorithm is used to fuse the image data in the frequency domain to eliminate the brightness difference and color difference between the images, and obtain a seamlessly stitched panoramic image.

6. The open-pit mine disaster detection method based on drone according to claim 5 is characterized in that: When collecting raw image data, the exposure parameters of the onboard optical sensor are controlled to obtain the raw image data collected by the sensor, and the exposure time and gain coefficient of the sensor are dynamically adjusted according to the brightness distribution of the image data.

7. The open-pit mine disaster detection method based on drone according to claim 5 is characterized in that: The preprocessing method comprises: according to the intrinsic parameters and extrinsic parameters of the optical sensor, using a radial distortion correction model to perform radial distortion and tangential distortion elimination processing on the image data.

8. The open-pit mine disaster detection method based on drone according to claim 5 is characterized in that: The method for obtaining attribute information of a suspected collapse area comprises: The image is detected in real time through the pre-trained YOLOv5 target detection model. When a suspected collapse area is detected, the suspected collapse area is located through a deep learning algorithm to obtain the coordinate information of the suspected collapse area, and the geographical location is obtained based on the coordinate information. The edge of the located suspected collapse area is extracted through an image segmentation algorithm to obtain the contour information of the suspected collapse area; based on the contour information, the area size of the suspected collapse area is calculated using the polygon area calculation formula; the coordinate information, geographical location and area size are associated and stored to form structured collapse area attribute information.

Images