High and steep dangerous rock identification method based on unmanned aerial vehicle multi-view image
Through the multi-view imaging technology of the UAV, combined with DSM contour line feature analysis and map library similarity comparison, the problem of geological disaster identification of high steep shore slopes is solved, the classification and hierarchy of dangerous rock bodies is realized, and the geological system is provided. Technical support for disaster monitoring.
Patent Information
- Application Number
- CN202410930294.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-11
- Publication Date
- 2024-11-22
AI Technical Summary
In areas with dense vegetation and high slopes and steep banks in the alpine canyon area, it is very difficult to manually inspect geological disaster hazards. A single-view drone image is difficult to effectively capture the engineering geological characteristics of high steep shore slopes, and cannot meet the needs of geological disaster hazard detection.
UAV multi-view imaging technology is used to obtain high-precision images through three-view angles (top view, strabismus, front view), generate DSM contour picture, and combine the similarity comparison between feature analysis and the map library to realize the classification and classification of dangerous rock mass The risk level is judged and the evolution trend is analyzed through multiple periods of image comparison.
The classification and hierarchical judgment of dangerous rock mass that is difficult for humans to reach high steep slopes is achieved. It can investigate large-scale dangerous rock mass at the same time. The identification results are more accurate as the data of the map library increases. The comparisons at different periods can reflect the changes in dangerous rock mass. Trends, economical and rapid, and identification results are more accurate as the data of the map library increases.
Smart Images

Figure CN119007035A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a method for identifying high-steep and dangerous rocks based on multi-view images of unmanned aerial vehicles. Background Art
[0002] The high mountain valleys are densely vegetated, with high slopes and steep banks, making it very difficult to manually inspect for potential geological disasters. Nearly 70% of the dozens of major geological disasters that have occurred in recent years were located in areas covered with vegetation or inaccessible areas. These high and steep slopes are difficult to detect through manual inspections.
[0003] In recent years, with the rise of drone photogrammetry technology and its increasing sophistication, the generation of high-precision three-dimensional real-scene models through drone close-up photography technology has been effective in the early identification of geological disasters such as landslides and collapses. The refined drone close-up photography technology can clearly obtain the deformation precursors of potentially unstable slopes, especially deformation cracks in rock walls or slope surfaces that have obvious grayscale texture differences from the surrounding environment.
[0004] For canyon areas with high and steep slopes, the geological features of the high and steep slopes are captured based on single-perspective drone images. Many engineering geological features cannot be captured well, making it difficult to meet the needs of geological disaster risk inspections.
[0005] Therefore, this study uses the "three perspectives" of aerial, oblique and frontal photography by drones to construct a map of the development characteristics of engineering geological conditions on steep slopes. Based on the engineering geological characteristic map of the "three perspectives", the engineering geological condition combination of geological hazard development factors is carried out, and the optimal combination of characteristic maps of different geological hazard disaster modes in high and steep canyon areas is constructed to identify geological hazard hazards on steep slopes. Summary of the invention
[0006] The purpose of the present invention is to provide a method for identifying high-steep dangerous rocks based on multi-view images of unmanned aerial vehicles. Based on the high-precision images obtained by three-view photography of unmanned aerial vehicles, DSM contour line pictures are generated after correction. By comparing the similarity between the data obtained by feature analysis of the DSM contour lines and the feature data of the atlas library, the classification of dangerous rock bodies can be preliminarily realized. At the same time, the characteristic value analysis of the DSM contour lines can preliminarily determine the danger level of the dangerous rock bodies, providing technical support for engineering construction and long-term monitoring.
[0007] In order to solve the above problems, the technical solution of the present invention is: The method for identifying high-steep dangerous rocks based on multi-view images of unmanned aerial vehicles is characterized by comprising the following steps: Step 1: Using an unmanned aerial vehicle to perform three-view photography, and performing spatial correction on the image data to construct a three-view DOM model; Step 2: Establishing a DSM model based on the image data taken by the unmanned aerial vehicle, and extracting the facade contour lines; Step 3: Performing feature extraction on the facade contour lines of the DSM model; Step 4: Establishing a feature atlas library, connecting with the atlas library to identify dangerous rock mass elements and their positions, and performing element combination; Step 5: Rapidly identifying dangerous rock mass based on the element combination; Step 6: By comparing multiple periods of unmanned aerial vehicle image data, analyzing the evolution trend of dangerous rock mass, and forming a visualization of the evolution process.
[0008] The beneficial effects of the present invention are: The present invention is based on high-precision images taken from three perspectives by drones, and generates DSM contour images after correction. By comparing the similarity between the data obtained from the characteristic analysis of the DSM contours and the characteristic data of the atlas library, it can preliminarily realize the classification of dangerous rock bodies on high and steep slopes that are difficult for humans to reach. At the same time, the characteristic value analysis of the DSM contours can preliminarily determine the danger level of the dangerous rock bodies. At the same time, only pictures taken by drones and algorithms are needed to determine the type and level of dangerous rock bodies. Compared with traditional methods, it has the following advantages: 1. A large range of dangerous rock bodies can be investigated at the same time; 2. Only pictures taken by drones are needed to realize the identification and classification of dangerous rock bodies, which is economical and fast; 3. The identification results will be more accurate as the atlas library data increases; 4. The comparison of results in different periods can reflect the changing trend of dangerous rock bodies. BRIEF DESCRIPTION OF THE DRAWINGS
[0009] The present invention will be further described below in conjunction with the accompanying drawings: Figure 1 is a flow chart of the present invention, Figure 2 This is a diagram of the high-steep dangerous rock identification system of the present invention. Figure 3 This is an example of a through-unloading crack sample of the present invention. Figure 4 This is an example of a vertical slope crack sample of the present invention. Figure 5 This is an example of the slope fragmentation and block falling samples of the present invention. DETAILED DESCRIPTION
[0010] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0011] like Figures 1 to 5 As shown in the figure, the high-steep dangerous rock recognition method based on multi-view images of UAV includes the following steps: Step 1: Use a drone to take three-view photography, perform spatial correction on the image data, and build a three-view DOM model; Step 2: Establish a DSM model based on the image data taken by the drone and extract the facade contour lines; Step 3: Extract features of the elevation contour lines of the DSM model; Step 4: Connect with the atlas library to identify dangerous rock mass elements and their locations, and combine the elements; Step 5: Rapidly identify dangerous rock masses based on the combination of factors; Step 6: By comparing multiple periods of UAV image data, analyze the evolution trend of dangerous rock masses, visualize the evolution process, and then determine the change trend of dangerous rock mass elements.
[0012] In step 1, after the drone takes off, it flies steadily to the specified height and hovers in the air, waiting for the camera to adjust to the characteristic angle. It is aimed directly at the river bank slope and achieves the effect of three-view shooting. Ensure that the picture is clear and stable. The drone flies slowly along the river bank slope, maintains a stable flight altitude and speed, and adjusts its posture to ensure the integrity of the picture. During the flight, the overall appearance and characteristics of the river bank slope are captured to capture the best effect; In this way, the drone can take three-view photography including front view (erect view), top view (orthographic view) and oblique view (45° fitting surface), among which the front view image is an image obtained by processing the vertical fitting plane of the ground points of the selected area as the projection surface; the top view image is an image obtained by processing the orthographic space fitting plane of the ground points of the selected area as the projection surface; the oblique view image is an image obtained by using the orthographic space fitting plane of the ground points of the selected area and the camera lens at a 45° angle as the projection surface.
[0013] In step 1, the method for spatially correcting the captured bird's-eye view, oblique view and front view image data includes: Spatial correction of overhead images: It is to realize the transformation of image translation, scaling, rotation, reflection, etc. through the direct connection method or inverse solution method, and locate the original image to the specified standard coordinate system. At the same time, based on the ground elevation information model (DEM), considering the impact of ground undulation on each pixel, the collinear equation is used as a mathematical model for digital differential correction, and finally the overhead (orthophoto) image is obtained. Spatial correction of squint images: by obtaining the spatial fitting plane, resampling the ground points in the selected area with set specifications, obtaining the sparse ground points in the selected area, and calculating the spatial fitting plane through spatial plane fitting, and obtaining the normal vector of the spatial fitting plane. The coordinate system is rotated to the 45° best fitting plane, and the coordinate system of the entire data is rotated according to the angular relationship between the plane normal vector in the space and the X, Y, and Z in the plane rectangular coordinate system, so that the obtained spatial fitting plane is parallel to the XOY plane of the umbrella mark system, thereby obtaining a squint (45° best fitting plane) image.
[0014] Spatial correction of the orthographic image: It is achieved by obtaining the spatial stereoscopic fitting plane, rotating the coordinate system to a 90° fitting plane, and rotating the coordinate system of the entire data according to the angular relationship between the plane normal vector in the space and the X, Y, and Z plane rectangular coordinate system, so that the obtained spatial fitting plane is perpendicular to the XOY plane of the parachute system, thereby obtaining the orthographic image.
[0015] In step 2, a DSM model is established based on the drone image, and the method for extracting features of the elevation contour lines of the DSM model includes: A: Use the software that can calculate contour lines (e.g. Pix4D) on your computer to import the corrected three-view images; B: Mark control points and check points in the image to check accuracy; C: Process the image, including image registration, image matching, 3D model reconstruction, etc.; D: Generate DSM model and edit and optimize it; E: Extract the facade contour lines according to the generated DSM model, and set the contour line type and color spacing and other parameters; F: Generate contour lines based on the generated DSM model, set contour line type and color spacing and other parameters, and output contour map.
[0016] In step 3, the method for extracting features from the DSM contour image includes: 1) using the edge detection algorithm in the computer, such as the Canny operator, first graying the contour image, then using the Canny operator to detect edges, and then extracting shape features and spacing features from the detected edges. 2) Shape description algorithm: first calculate the Zernike moment of the image, and then use the Zernike moment to describe the shape features of the image.
[0017] In step 4, the atlas library is established based on the characteristics of high-steep dangerous rock development from three perspectives, and the establishment method includes the following steps: Step 1: Conduct basic feature sample library training on the locations of relevant characteristic elements such as the boundaries, bases, rock formation occurrence, structural surfaces, fracture development, and rock mass degradation of high-steep and dangerous rocks; Step 2: Extract features of the key contour lines of each type of element, identify and extract key information, such as the shape, texture, spacing, height difference, etc. of the contour lines; Step 3: Describe and encode the extracted contour features; Step 4: Store the extracted and described contour feature information in the atlas library.
[0018] In the third step above, the method for describing and encoding the extracted contour features includes: SA1, extract features from each contour image, including shape features and spacing features; SA2, stores the extracted feature information as a data structure (such as a dictionary or data frame); SA3 integrates and summarizes the feature information extracted from all contour images, and calculates statistics such as the mean and standard deviation. Further processing such as feature selection and dimensionality reduction can be performed as needed. Finally, the integrated and summarized feature information is output.
[0019] In the fourth step, the method for storing the extracted and described contour feature information in the atlas library comprises the following steps: SB1, create a database table to store the feature information of the contour image of the feature type A: store the feature information extracted from each image, including shape features and spacing features, etc.; SB2, write a Python script or program on your computer to read the feature information of each contour image and insert it into the Type A database table. You can use a library such as SQLAlchemy to connect to the database and perform the insert operation; SB3, repeat the above operation to establish different element type databases respectively; SB4, different types of databases are integrated into a map library.
[0020] In step 4, the dangerous rock mass elements and their locations are identified by docking with the atlas library, and the combined identification method includes the following steps: SC1, compare the newly extracted feature information with the feature information of type A database; SC2, use an appropriate algorithm (such as KNN) to calculate the similarity to determine whether the new image contains contour lines of type A. Based on the comparison results, calculate the number of contour lines of type A and express the results; SC3, repeat the above operation, make similarity judgment with the feature values of different databases, and judge how many types of contour lines the contour lines in the area have; SC4, integrated output results, the results include contour line type and location; SC5, classify and combine contour line types and locations.
[0021] In step five, the method for determining the type of dangerous rock mass based on the combination of factors is as follows: classify and combine the contour line type and location of each type of dangerous rock mass, formulate identification rules, make one-to-one correspondence between the combination of factors and the type of dangerous rock mass, and determine the type of dangerous rock mass based on the identified combination of factors.
[0022] In step five, the method for judging the dangerous rock mass level according to the combination of factors is as follows: setting intervals for different types of contour line characteristic values and classifying them, and different intervals correspond to different danger levels.
[0023] In step six, the method of comparing multiple images and analyzing the evolution trend of dangerous rock masses includes: after processing images of different periods, establishing a DSM model and generating contour images, extracting contour features from the contour images, comparing the type changes of contour lines in the same area, and outputting the change results if the type does not change. If the type does not change, comparing the changes in characteristic values, such as the spacing, and outputting the level changes according to the corresponding dangerous interval. If there is no change, no result is output.
[0024] In step 6, the steps of the visualization method for comparing multiple images and analyzing the deformation trend of dangerous rock mass include: SD1, save the results in a spatial data format, such as Shapefile or GeoJSON. At the same time, ensure that the contour data of different periods are in the same coordinate system; SD2, use the GIS software in the computer to load the contour data and feature data of different periods, set the style of the contour data and feature data of different periods, and set the color gradient or symbol size according to the size of the feature value; SD3, superimposes contour data and feature data from different periods on the same map to compare feature changes; SD4, use the layout function of GIS software to create map layout. Export the map as an image or PDF file to show the spatial distribution of feature changes.
[0025] The contents described in the embodiments of this specification are merely an enumeration of the implementation forms of the inventive concept. The protection scope of the present invention should not be regarded as limited to the specific forms described in the embodiments. The protection scope of the present invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.
Claims
1. A method for identifying high-steep and dangerous rocks based on multi-view images from unmanned aerial vehicles, characterized by: The steps include: Step 1: Use a drone to take three-view photography, perform spatial correction on the image data, and build a three-view DOM model; Step 2: Establish a DSM model based on the image data taken by the drone and extract the facade contour lines; Step 3: Extract features of the elevation contour lines of the DSM model; Step 4: Establish a feature atlas library, connect with the atlas library to identify dangerous rock mass elements and their locations, and combine the elements; Step 5: Rapidly identify dangerous rock masses based on the combination of factors; Step 6: By comparing multiple periods of UAV image data, analyze the evolution trend of dangerous rock masses and visualize the evolution process.
2. The method for identifying high-steep and dangerous rocks based on multi-view images from unmanned aerial vehicles according to claim 1 is characterized by: In step 1, the three-view photography of the drone includes a bird's-eye view, an oblique view, and a front view. The method for spatially correcting the bird's-eye view, oblique view, and front view image data includes: The spatial correction of the overhead image is to achieve the image translation, scaling, rotation, reflection and other transformations through the direct method or inverse method, locate the original image to the specified standard coordinate system, and use the collinear equation as the mathematical model based on the ground elevation DSM model to perform digital differential correction, and finally obtain the overhead image; The spatial correction of the strabismus image is to obtain the spatial fitting plane, resample the ground points in the selected area, obtain the sparse ground points in the selected area, calculate the spatial fitting plane through spatial plane fitting, and obtain the normal vector of the spatial fitting plane, and convert the coordinate system to the fitting plane. According to the angular relationship between the plane normal vector in the space and the X, Y, and Z in the plane rectangular coordinate system, the coordinate system of the entire data is rotated so that the obtained spatial fitting plane is parallel to the XOY plane of the umbrella mark system, thereby obtaining the strabismus image; The spatial correction of the orthographic image is achieved by obtaining the spatial stereoscopic fitting plane, rotating the coordinate system to a 90° fitting plane, and rotating the coordinate system of the entire data according to the angular relationship between the plane normal vector in the space and the X, Y, and Z plane rectangular coordinate system, so that the obtained spatial fitting plane is perpendicular to the XOY plane of the parachute system, thereby obtaining the orthographic image.
3. The method for identifying high-steep and dangerous rocks based on multi-view images from unmanned aerial vehicles according to claim 1 is characterized by: In step 2, the method for extracting features of the elevation contour lines of the DSM model includes: importing the corrected three-view images into a computer, generating and editing a DSM model on the computer, extracting elevation contour lines based on the generated DSM model, and extracting the development characteristics of the slope rock and soil based on the contour line features.
4. The method for identifying high-steep and dangerous rocks based on multi-view images from unmanned aerial vehicles according to claim 3 is characterized by: The method for extracting the development characteristics of the rock and soil mass on the bank slope includes: firstly using the edge detection algorithm to grayscale and detect the edges of the contour image, and then extracting the shape features and spacing features of the detected edges; finally, using the shape description algorithm to calculate the Zernike moment of the image, and then using the Zernike moment to describe the shape characteristics of the image.
5. The method for identifying high-steep and dangerous rocks based on multi-view images from unmanned aerial vehicles according to claim 1 is characterized by: In step 4, the atlas library is established based on the characteristics of high-steep dangerous rock development from three perspectives, and the establishment method includes the following steps: Step 1: Conduct basic feature sample library training on the locations of characteristic elements of high-steep and dangerous rocks; Step 2: Extract features of key contour lines of each type of element, identify and extract key information; Step 3: Describe and encode the extracted contour features; Step 4: Store the extracted and described contour feature information in the atlas library.
6. The method for identifying high-steep and dangerous rocks based on multi-view images from unmanned aerial vehicles according to claim 5 is characterized by: In the third step, the method of describing and encoding the extracted contour features includes: SA1, extract features from each contour image, including shape features and spacing features; SA2, stores the extracted feature information as a data structure; SA3 integrates and summarizes the feature information extracted from all contour images, calculates statistics such as the mean and standard deviation, and finally outputs the integrated and summarized feature information.
7. The method for identifying high-steep and dangerous rocks based on multi-view images from unmanned aerial vehicles according to claim 5 is characterized by: In the fourth step, the method for storing the extracted and described contour feature information in the atlas library includes the following steps: SB1, create a database table to store the feature information of the contour image of the feature type A: store the feature information extracted from each image, including shape features and spacing features; SB2, write a Python script or program on your computer to read the feature information of each contour image and insert it into the Type A database table, using the SQLAlchemy library to connect to the database and perform the insert operation; SB3, repeat the above operation to establish different element type databases respectively; SB4, different types of databases are integrated into a map library.
8. The method for identifying high-steep and dangerous rocks based on multi-view images from unmanned aerial vehicles according to claim 1 is characterized by: In step 4, the dangerous rock mass elements and their locations are identified by docking with the atlas library, and the combined identification method includes the following steps: SC1, compare the newly extracted feature information with the feature information of type A database; SC2, use the algorithm to calculate the similarity to determine whether the new image contains type A contour lines. Based on the comparison results, calculate the number of type A contour lines and express the results. SC3, repeat the above operation, make similarity judgment with the feature values of different databases, and judge how many types of contour lines the contour lines in the area have; SC4, integrated output results, the results include contour line type and location; SC5, classify and combine contour line types and locations.
9. The method for identifying high-steep and dangerous rocks based on multi-view images from unmanned aerial vehicles according to claim 1 is characterized by: In step six, the method of comparing multiple images and analyzing the evolution trend of dangerous rock masses includes: after processing images of different periods, establishing a DSM model and generating contour images, extracting contour features from the contour images, comparing the type changes of contour lines in the same area, and outputting the change results if the type does not change. If the type does not change, then comparing the changes in the characteristic values, and outputting the level changes according to the corresponding dangerous interval. If there is no change, no result is output.
10. The method for identifying high-steep and dangerous rocks based on multi-view images from unmanned aerial vehicles according to claim 1, characterized in that: In step 6, the visualization method of comparing multiple images and analyzing the deformation trend of dangerous rock mass includes the following steps: SD1, save the results in spatial data format, and ensure that the contour data of different periods are in the same coordinate system; SD2, the computer loads the contour data and feature data of different periods, sets the styles of the contour data and feature data of different periods, and sets the color gradient or symbol size according to the size of the feature value; SD3, superimposes contour data and feature data from different periods on the same map; SD4, using the layout function of the computer, creates a map layout, exports the map as an image or PDF file, and displays the spatial distribution of feature changes.
Citation Information
Patent Citations
Visual image monitoring and recognition system
CN111861982A
Aerial survey method and system based on real-time dense three-dimensional point cloud of unmanned aerial vehicle and DSM
CN112434709A
Fragmented and loosened rock mass development range delineation method based on three-dimensional space image technology
CN114898053A
High and steep slope dangerous rock mass rapid identification method based on unmanned aerial vehicle LiDAR terrain simulation flight
CN115439762A
Method for identifying potential landslide hazard of reservoir bank based on rock mass degradation feature
US20220383475A1
