CN-121980429-A - Debris flow object source monitoring method and system

CN121980429ACN 121980429 ACN121980429 ACN 121980429ACN-121980429-A

Abstract

According to the debris flow object source monitoring method and system, feature integration processing is carried out on the data monitoring features corresponding to the debris flow attribute data according to the matching data between the data monitoring features corresponding to the debris flow attribute data, and the global data monitoring features corresponding to the debris flow object source data set to be identified are obtained. And analyzing and processing the global data monitoring characteristics to obtain an analysis result corresponding to the debris flow source data set to be identified. In the analysis processing task of the object source data set, the data monitoring characteristics corresponding to a plurality of debris flow attribute data in the object source data set of the debris flow to be identified can be combined to obtain global data monitoring characteristics with better characterization force, the object source hidden danger analysis result can be accurately obtained through the global data monitoring characteristics, and the analysis processing accuracy of the object source data set can be improved.

Inventors

YANG XUEZHI
LIU KANGLIN
MENG MINGHUI
SUN DONG
LU SHUAI
ZOU XUEQING
TANG LIANG

Assignees

四川省华地建设工程有限责任公司
四川省地质环境调查研究中心

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (11)

1. The object source monitoring method for the debris flow is characterized by comprising the following steps of: And obtaining X debris flow attribute data corresponding to the debris flow object source data set to be identified, wherein the X debris flow attribute data comprise six categories of object source types, spatial distribution, geometric forms, physical mechanics, reserves and replenishment and dynamic evolution.
2. Extracting characteristics of each debris flow attribute data to obtain data monitoring characteristics corresponding to each debris flow attribute data, and obtaining debris flow object source attribute matching data between the data monitoring characteristics corresponding to each debris flow attribute data; Performing feature integration processing on the data monitoring features corresponding to each debris flow attribute data based on the debris flow source attribute matching data to obtain global data monitoring features corresponding to the debris flow source data set to be identified; and analyzing and processing the global data monitoring characteristics to obtain an analysis result corresponding to the debris flow source data set to be identified.
3. The method of claim 1, wherein the obtaining X debris flow attribute data corresponding to the debris flow source data set to be identified comprises: Acquiring a debris flow object source attribute type range in a debris flow object source monitoring instruction, and analyzing the debris flow object source data set to be identified based on the debris flow object source attribute type in the debris flow object source attribute type range to obtain original data matched with the debris flow object source attribute type in the debris flow object source attribute type range; and carrying out data filtering processing on the original data to obtain X debris flow attribute data corresponding to the debris flow object source data set to be identified.
4. The method of claim 2, wherein the performing data filtering on the raw data to obtain X debris flow attribute data corresponding to the debris flow source data set to be identified includes: performing quality verification on mountain item information in the original data to obtain a mountain item quality verification result corresponding to the mountain item information; If the mountain item quality verification result indicates that the verification is qualified, carrying out standardized processing on the mountain item information to obtain debris flow source hidden danger standard data; obtaining key material source type data from material source type data covered by the original data, and performing interference elimination processing on material source historical data in the original data to obtain interference elimination material source historical data; And obtaining X debris flow attribute data corresponding to the debris flow object source data set to be identified based on the debris flow object source hidden danger standard data, the key object source type data and the interference elimination object source historical data.
5. The method of claim 1, wherein the X debris flow attribute data includes key source type data, interference elimination source history data and debris flow source hidden danger standard data, wherein the feature extraction is performed on each debris flow attribute data to obtain data monitoring features corresponding to each debris flow attribute data, and the method comprises the steps of: converting the key object source type data into object source weight characteristics, carrying out convolution processing on the object source weight characteristics to obtain hidden danger weight characteristics, and taking the object source weight characteristics and the hidden danger weight characteristics as data monitoring characteristics corresponding to the key object source type data; Performing convolution processing on the interference-removed object source historical data to obtain object source factor characteristics corresponding to the interference-removed object source historical data; Analyzing the interference-free object source historical data to obtain a characteristic analysis result, and taking the object source factor characteristic and the characteristic analysis result as data monitoring characteristics corresponding to the interference-free object source historical data; And obtaining confidence coefficient integration information associated with the debris flow source hidden danger standard data, and carrying out convolution processing on the confidence coefficient integration information and the debris flow source hidden danger standard data to obtain data monitoring characteristics corresponding to the debris flow source hidden danger standard data.
6. The method of claim 4, wherein said converting said key source species data into source weight features comprises: Converting the key object source type data into W hidden danger factors to obtain hidden danger vectors corresponding to the W hidden danger factors, wherein the W hidden danger factors comprise real-time rainfall hidden danger factors, deposit hidden danger factors and blocking hidden danger factors; Based on important information of the W hidden danger factors in the key object source type data, obtaining feature vectors corresponding to the W hidden danger factors; positioning hidden danger areas in the key object source type data based on the W hidden danger factors to obtain positioning vectors corresponding to the W hidden danger factors; And combining the hidden danger vector, the feature vector and the positioning vector to obtain the object source weight feature corresponding to the key object source type data.
7. The method of claim 4, wherein the convolving the de-interfering object source history data to obtain object source factor characteristics corresponding to the de-interfering object source history data, comprises: The interference removing object source historical data is loaded to an object source data convolution unit, and the input characteristic of an mth characteristic extraction module in the object source data convolution unit is obtained, wherein when m is 1, the input characteristic of the mth characteristic extraction module is the interference removing object source historical data; processing the input features of the mth feature extraction module based on one or more feature extraction layers in the mth feature extraction module to obtain a processing result; Based on the weight coefficient corresponding to the simplified layer in the mth feature extraction module, simplifying the processing result to obtain simplified features; And combining the simplified features with the input features of the mth feature extraction module to obtain the output features of the mth feature extraction module, and taking the output features of the Y feature extraction module in the object source data convolution unit as the object source factor features corresponding to the interference removing object source historical data.
8. The method of claim 4, wherein analyzing the de-interfering source history data to obtain a signature analysis result comprises: performing data detection on the interference-free object source historical data to obtain a data detection result, and cleaning the interference-free object source historical data based on the data detection result to obtain a cleaning result; and classifying the cleaning result to obtain a classification result, and analyzing the classification result to obtain a characteristic analysis result corresponding to the interference removing object source historical data.
9. The method of claim 1, wherein the performing feature integration processing on the data monitoring feature corresponding to each of the debris flow attribute data based on the debris flow source attribute matching data to obtain the global data monitoring feature corresponding to the debris flow source data set to be identified includes: If the debris flow object source attribute matching data indicates the data monitoring feature corresponding to the a-th debris flow attribute data in the X debris flow attribute data and the data monitoring feature corresponding to the b-th debris flow attribute data have an association relationship, the debris flow attribute data corresponding to the a-th debris flow attribute data and the data monitoring feature corresponding to the b-th debris flow attribute data are fused to obtain a debris flow object source attribute combination feature; Processing the debris flow source attribute combination characteristics and the data monitoring characteristics corresponding to each debris flow attribute data to obtain global data monitoring characteristics corresponding to the debris flow source data set to be identified; The analyzing the global data monitoring feature to obtain an analysis result corresponding to the debris flow source data set to be identified comprises the following steps: Loading the global data monitoring characteristics to a first data analysis network, and outputting a first regression analysis result corresponding to the debris flow source data set to be identified through the first data analysis network; loading the global data monitoring characteristics to a second data analysis network, and outputting a second regression analysis result corresponding to the debris flow source data set to be identified through the second data analysis network; And performing function processing on the first regression analysis result and the second regression analysis result to obtain an analysis result corresponding to the debris flow source data set to be identified.
10. The method of any one of claims 1 to 8, further comprising: Obtaining X sample debris flow attribute data corresponding to the sample object source data set; Feature extraction is carried out on each sample debris flow attribute data to obtain sample data monitoring features corresponding to each sample debris flow attribute data, and feature matching data among the sample data monitoring features corresponding to each sample debris flow attribute data is obtained; Based on the feature matching data, performing feature integration processing on the sample data monitoring features corresponding to each sample debris flow attribute data to obtain global debris flow source attribute sample features corresponding to the sample source data set; Loading the global debris flow source attribute example features to a pre-training network, and analyzing and processing the global debris flow source attribute example features through the pre-training network to obtain example categories corresponding to the example source data sets; Optimizing network coefficients of the pre-training network based on the abnormality between the example category and the example catalog corresponding to the example source data set, and taking the pre-training network covering the optimized network coefficients as an object source data set data analysis network, wherein the object source data set data analysis network comprises at least one of a first data analysis network and a second data analysis network.
11. A debris flow source monitoring system comprising a processor and a memory in communication with each other, the processor being adapted to read a computer program from the memory and execute the computer program to implement the method of any one of claims 1-9.

Description

Debris flow object source monitoring method and system Technical Field The application relates to the technical field of debris flow object source monitoring, in particular to a debris flow object source monitoring method and system. Background The debris flow is used as a mountain area high-occurrence geological disaster, the formation of the debris flow is closely related to the occurrence state, stability and activation condition of the object source, the occurrence probability, scale and hazard degree of the debris flow are directly determined by the type, reserve, distribution and dynamic evolution characteristics of the object source, so that the debris flow is accurately and efficiently monitored and analyzed, and the debris flow is a core premise for realizing early warning and scientific prevention and control of the debris flow disaster. At present, the debris flow object source monitoring field has formed a monitoring system based on various technologies such as site survey, remote sensing interpretation, sensor monitoring and the like, and can collect multi-dimensional original data such as object source types, spatial distribution, physical and mechanical properties, dynamic evolution and the like, but a plurality of technical short boards still exist in the data processing and analysis links, so that the requirements of fine and intelligent object source monitoring are difficult to meet, and the specific problems are as follows: The data processing lacks a standardized system, namely, the original data of the obtained debris flow source is monitored to be various in sources, heterogeneous in format and complicated in dimension, and contains a large amount of invalid interference data, the prior art does not establish a unified filtering and standardized processing flow aiming at multi-attribute data of the source, the data is easy to be low in effectiveness and poor in consistency, a reliable data basis cannot be provided for subsequent analysis, and meanwhile, a reasonable complement strategy is lacking aiming at the missing attribute data of the source, so that the comprehensiveness of monitoring and analysis is further influenced. The characteristic extraction method is single and has the defects of obvious intrinsic characteristic rule difference of different attribute data of the object source, the prior art mostly adopts a general characteristic extraction method to process the object source data, and a differentiated characteristic extraction strategy is not designed according to the characteristics of different types of data such as key object source type data, historical evolution data, hidden danger standard data and the like, so that core monitoring characteristics contained in the data are difficult to mine, the characteristic representation capability is weak, and the actual hidden danger state of the object source cannot be accurately reflected. The characteristic integration lacks relevance consideration that the prior art integrates the extracted object source characteristics by adopting a simple stacking or splicing mode, does not analyze the internal relevance between the attribute characteristics of different object sources, is easy to cause characteristic redundancy and information overlapping, cannot construct global characteristics capable of comprehensively reflecting the overall characteristics of the object sources, ensures that the subsequent analysis can only reflect the state of single dimension of the object sources, and is difficult to realize the overall judgment of hidden danger of the object sources. The analysis model has low precision and poor generalization capability, the existing debris flow source analysis depends on a manual interpretation or simple machine learning model, the manual participation degree is high, the efficiency is low and the subjectivity is strong, the traditional model is mostly based on single-dimension feature training, the multi-attribute source features are not fully fused, the model training lacks a special optimization flow aiming at debris flow source scenes, the identification accuracy of the model on the hidden danger of the debris flow source is low, the generalization capability is weak, and the monitoring and analysis requirements of the debris flow sources of different drainage basins and different types are difficult to adapt. The intelligent and automatic degree of monitoring analysis is low, the prior art does not build a full-flow intelligent technical system from the original data acquisition, feature extraction and feature integration to analysis result output, and the links lack effective connection, so that the data screening, feature processing and result interpretation are finished by a large amount of manual intervention, the monitoring operation cost is increased, the analysis efficiency is low, and the normalized and real-time monitoring of the debris flow source cannot be realized. In summary, aiming at the d