CN-122020562-A - Small and medium-sized high-order collapse geological disaster hidden danger identification method and device based on micro-topography characteristics and storage medium

CN122020562ACN 122020562 ACN122020562 ACN 122020562ACN-122020562-A

Abstract

The application provides a method, a device and a storage medium for identifying hidden danger of a small and medium-sized high-order collapse geological disaster based on micro-topography features, and relates to the technical field of geological disaster hidden danger identification. The identification method comprises the steps of obtaining an easy-disaster micro-terrain characteristic index system of a slope unit to be identified in a working area, obtaining a plurality of first-level indexes including the shape of the slope unit, inputting a multi-dimensional geometric characteristic vector formed based on the second-level indexes after standardization processing into a trained hidden danger identification model, outputting a reconstruction characteristic vector identical to the input dimension by the hidden danger identification model, calculating a reconstruction error by comparing the input and output differences, obtaining a potential slump index based on the weighted fusion of the reconstruction error and the second-level indexes, and enabling the larger numerical value to represent the higher possibility of slump disasters of the slope. By adopting the identification method provided by the application, the hidden danger of small and medium-sized high-order collapse of the high-risk area of the geological disaster can be efficiently and accurately identified.

Inventors

TONG BIN
LI YUANHUA
Sa Lanpeng
XU HONGTAO
ZHANG NAN
He Jianyin
YIN YUEPING
LIU YINGJUN
Tang Jiting
LI BING
Lei Dewen
ZHANG WEIFENG
ZHU CHUANBING
ZHANG YONG

Assignees

中国地质环境监测院(自然资源部地质灾害技术指导中心)
中国矿业大学(北京)
云南省地质环境监测院(云南省环境地质研究院)
云南地质工程勘察设计研究院有限公司
云南省地质工程勘察有限公司
北方国际合作股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260212

Claims (10)

1. The method for identifying the hidden danger of the small and medium-sized high-level collapse geological disaster based on the micro-topography features is characterized by comprising the following steps of: Acquiring a disaster-prone micro-topography characteristic index system of a slope unit to be identified in a working area, wherein the disaster-prone micro-topography characteristic index system comprises a plurality of primary indexes including a slope unit form, a boot-shaped slope form, a slope turning belt, a slope surface concave cavity, a slope vertical empty face and a micro-topography combined form, and each primary index comprises a plurality of secondary indexes; inputting a multidimensional geometric feature vector formed based on each secondary index after standardization processing into a trained hidden danger identification model, outputting a reconstructed feature vector with the same dimension as the input hidden danger identification model, and calculating a reconstruction error by comparing the input and output differences; And obtaining potential collapse indexes of the slope unit to be identified according to the reconstruction errors and introducing weighted fusion of the secondary indexes, wherein the larger the potential collapse index value is, the higher the possibility of collapse disasters on the slope is.
2. The identification method according to claim 1, wherein each primary index comprises a plurality of secondary indexes, specifically: The secondary indexes for calculating and extracting the shape of the slope unit according to the three-dimensional geometrical characteristics of the section line of the slope unit comprise the length of the slope of the main section, the width of the slope of the main section, the length-width ratio of the slope unit along the direction of the main section, the total height difference from the topmost part to the bottommost part of the slope, the length-width ratio, the area under the section curve, the area ratio under the section line and the area ratio; Identifying gradient turning points according to three-dimensional geometric characteristics of a slope section line, dividing the slope into a steep bank area and a gentle slope area, and calculating and extracting secondary indexes of the boot-shaped slope form, wherein the secondary indexes comprise a boot-shaped terrain judging value, a steep bank area length, a steep bank area height, a gentle slope area length, a gentle slope area height, a steep bank area height ratio, a steep bank area gradient and a gentle slope area gradient; Calculating the turning belt width of the waistcoat turning belt, extracting the height difference along the waistcoat line, and calculating the turning belt height difference of the waistcoat turning belt; a DEM model generated based on three-dimensional point cloud data is used for extracting triangular surfaces with downward directions, cavities are obtained through regional growth, and the secondary indexes of the surface cavities of the slope surface are calculated and extracted, wherein the secondary indexes of the surface cavities of the slope surface comprise the number of the cavities, the depth of the cavities, the area of the cavities, the height of the lower edge of the cavities, the section curvature of the cavities and the plane curvature of the cavities; A DEM model generated based on three-dimensional point cloud data is used for designating a normal vector direction and an included angle threshold value, an upright empty face is obtained through communication analysis, and the calculated and extracted secondary indexes of the upright empty face of the slope body comprise the number of the upright empty faces, the gradient of the empty face, the height of the empty face, the area of the empty face, the upper edge height of the empty face and the slope change rate; Based on space superposition analysis, calculating the shortest Euclidean distance between the lower edge of the concave cavity and the boundary of the temporary face to obtain the concave cavity-temporary face distance of the micro-topography combined form, extracting the total area of the concave cavity and the temporary face in the abrupt bank area, and calculating the ratio of the total area of the concave cavity and the temporary face to the area of the abrupt bank area to obtain the combined area ratio of the micro-topography combined form.
3. The identification method according to claim 2, wherein the identified gradient turning points specifically refer to: Calculating the curvature of each point of the section line Traversing all points to obtain the lowest local curvature value, i.e. concave point on the section, if there is only one concave point, the obtained concave point is the turning point of abrupt bank zone and gentle slope zone, if there are several concave points, selecting one of the points with farthest vertical distance to the connecting line of starting point and end point, and/or Determining a concave cavity range by adopting a region growing method, namely acquiring any one vertex which is normal downwards as a seed, adding a queue and marking the accessed vertex, outwards growing a neighborhood according to a breadth-first principle, expanding and adding the adjacent point into the queue if the normal vector included angle between the adjacent point and the seed point is smaller than a threshold value and the adjacent point is also a concave point, stopping expanding if all the adjacent vertexes do not meet the above condition or reach a Mesh boundary, judging that the concave cavity range obtained by growth is larger than a certain threshold value as an effective concave cavity, and/or The method for obtaining the vertical empty face through communication analysis specifically comprises the steps of calculating a normal vector of each triangular face, filtering through an included angle threshold value between the normal vector and a horizontal plane to obtain triangular faces with gradient meeting conditions, conducting communication analysis on the triangular faces meeting the conditions to obtain one or more communication areas, calculating the area of the communication areas and the average curvature of all vertexes, wherein the area is larger than a specified threshold value, and the absolute value of the average curvature is smaller than the specified threshold value, and taking the communication areas as the empty vertical face.
4. The identification method according to claim 1, wherein the normalization process specifically refers to: Each secondary index is standardized, the stability of the model is ensured, In the formula, Is a two-level index which is used for the two-level index, Is the mean value of the ith quantization index, Is the standard deviation of the ith quantization index.
5. The recognition method according to claim 1, wherein the hidden danger recognition model outputs a reconstructed feature vector identical to the input dimension, and the reconstruction error is calculated by comparing the difference between the input and the output, specifically: The hidden danger identification model is constructed by adopting a self-encoder in the deep learning, the core structure of the hidden danger identification model comprises an encoder f and a decoder g, the encoder compresses an input x into a potential representation z, and the decoder reconstructs the input ; Wherein W 1 、W 2 is a weight matrix, b 1 、b 2 is a bias term, sigma is an activation function ReLU, and complex feature conversion is realized; The reconstruction error is calculated and the reconstruction error is calculated, In the formula, E is a reconstruction error, and the larger E is, the more abnormal the input characteristic is.
6. The method of claim 1, further comprising training the hidden danger identification model to obtain a trained hidden danger identification model, wherein training the hidden danger identification model comprises: The method comprises the steps of obtaining a disaster-prone micro-topography characteristic index system sample of a medium-sized and small-sized high-order collapse disaster slope unit sample in a working area, taking a sample set of multi-dimensional geometric characteristic vectors formed based on all disaster-prone micro-topography characteristic index system samples as a training set, and inputting the training set into an initial self-encoder model; An encoder and a decoder of the initial self-encoder model convert each geometrical feature vector of the training set into a reconstructed feature vector of the same dimension; The self-encoder model is used for determining the hidden danger identification model to be trained when the reconstruction error does not descend or starts ascending in a plurality of continuous rounds or the training model reaches the preset maximum training round number.
7. The identification method according to claim 1, wherein the potential slump index of the slope unit to be identified is obtained by weighting and fusing the reconstruction errors and introducing the secondary indexes, specifically: Defining a weight W i of each quantization index, wherein the weight is determined through statistical significance of historical samples and/or expert experience; Where I is the combination of the weighted sum and the reconstruction error, And For the reconciliation parameters, controlling the contribution of the weighted sum and the error term by training; the identification method further comprises normalizing the potential slump index, In the formula, Normalized index of potential collapse index, I can be normalized to [0,1], with a closer index to 1 indicating a higher risk of high-risk collapse.
8. The method of claim 7, further comprising classifying the potential slump index into three risk classes, namely high, medium and low: The risk levels are classified according to the corresponding relation between the potential collapse indexes and the set threshold value, wherein the potential collapse index is less than or equal to the low risk threshold value and is a high risk level, the potential collapse index is a medium risk level between the low and high threshold values, and the potential collapse index is greater than the high risk threshold value and is a low risk level; The judging threshold values comprise an area ratio judging threshold value, a concave cavity range judging threshold value, a normal vector included angle threshold value, an upright empty face area threshold value, an upright empty face average curvature threshold value, a potential collapse index high-risk judging threshold value and a reconstruction error low/high threshold value, and all the judging threshold values are calibrated and determined according to actual geological conditions and historical collapse case data in a working area.
9. Small and medium-sized high-order collapse geological disaster hidden danger identification device based on micro-topography features, characterized in that, identification device includes: a memory for storing computer executable instructions; a processor for implementing the identification method of any one of claims 1 to 8 when executing computer executable instructions stored in said memory.
10. A storage medium storing computer instructions for causing a computer to perform the identification method according to any one of claims 1 to 8.

Description

Small and medium-sized high-order collapse geological disaster hidden danger identification method and device based on micro-topography characteristics and storage medium Technical Field The application relates to the technical field of geological disaster hidden danger identification, in particular to a method and a device for identifying small and medium-sized high-order collapse geological disaster hidden danger based on micro-topography characteristics and a storage medium Background The medium-sized and small-sized high-level collapse is a typical geological disaster in China, is frequently generated at the middle and upper parts of steep slopes, has the characteristics of strong concealment, high burst performance, random spatial distribution and large impact destructive power, has high hidden danger identification difficulty, and is important and difficult for preventing and controlling the geological disaster. The collapse source area is far away from the toe and has high potential energy, so that the collapse disaster is difficult to identify by the traditional means, and life and property safety are seriously threatened. The medium and small high-order collapse disaster shows multiple specificities: firstly, the concealment is strong, a collapse source area is usually positioned on a high-level cliff which is difficult to directly observe by human activities, and the recognition difficulty of naked eyes and traditional investigation means is further increased by vegetation coverage or terrain shielding; Secondly, the burst is remarkable, the loosening of the rock mass to the final instability is often completed in only a few minutes to a few hours, and the precursor is difficult to effectively capture by the conventional monitoring means; and thirdly, the energy is concentrated and the impact range is large, although the collapse square quantity is generally between hundreds and tens of thousands of cubic meters and the square quantity is smaller, the high potential energy is converted into huge kinetic energy, so that impact-chip flow is often formed, and destructive striking is caused to linear engineering such as houses, roads and the like at the lower part. Currently, research and development of a large-scale, high-precision and intelligent hidden danger judging and identifying technology is one of the key points, difficulties and global neck problems in the field. The development of observation technologies such as unmanned aerial vehicle aerial survey and laser radar realizes the rapid acquisition of high-precision DEM data on an area, provides solid support for the extraction and quantification of slope micro-landform features, and the topography features are the most critical control factors for the occurrence of such disasters. However, the existing identification method pays attention to macro indexes, ignores the indication effect of micro-landforms, has the defects of incomplete feature selection, unreasonable weight distribution, realization of binary judgment and the like, and cannot accurately quantify hidden trouble degrees. Therefore, a recognition method capable of systematically extracting micro-geomorphic features, accurately quantifying hidden danger levels, efficiently and effectively and feasibility is needed, the defects of the prior art are overcome, and the accurate prevention and control requirements of geological disasters are met. Disclosure of Invention In view of the above, the embodiment of the application provides a method, a device and a storage medium for identifying hidden dangers of small and medium-sized high-order collapse geological disasters based on micro-topography features, which are used for at least solving one of the problems in the prior art. In a first aspect, an embodiment of the present application provides a method for identifying hidden danger of a small and medium sized high-order collapse geological disaster based on micro-topography features, where the identification method includes: Acquiring a disaster-prone micro-topography characteristic index system of a slope unit to be identified in a working area, wherein the disaster-prone micro-topography characteristic index system comprises a plurality of primary indexes including a slope unit form, a boot-shaped slope form, a slope turning belt, a slope surface concave cavity, a slope vertical empty face and a micro-topography combined form, and each primary index comprises a plurality of secondary indexes; inputting a multidimensional geometric feature vector formed based on each secondary index after standardization processing into a trained hidden danger identification model, outputting a reconstructed feature vector with the same dimension as the input hidden danger identification model, and calculating a reconstruction error by comparing the input and output differences; And obtaining potential collapse indexes of the slope unit to be identified according to the reconstruction errors and int