Development Of Highway Geological Alternatives Selection Method And System In Landslide Area

Geological alternatives selection is the core content of highway alternatives selection in mountainous areas.Research on geological alternatives selection methods based on GIS and intelligent technology can improve the decision-making level of highway alternatives selection.At present,geological alternatives selection mainly relies on expert experience,and lacks support for the whole process of corridor scheme formulation and comparison,design scheme comparison and selection models,and alignment knowledge reasoning.Aiming at the lack of quantitative and visual guidance information in the process of highway geological alternatives selection,the paper relies on the national key research and formulation plan "Multi-dimensional environmental factors and multi-objective intelligent line selection technology"(project number: 2021YFB2600403),taking the geological alternatives selection of landslides in the western Sichuan plateau as an example,this paper constructs a landslide susceptibility model based on machine learning,establishes a geological alternatives selection model framework based on landslide susceptibility,terrain reconstruction and knowledge reasoning,and designs and develops a geological alternatives selection system for highway landslide.The system realizes the decision-making information support in the whole process of geological alternatives selection.Collected and preprocessed the relevant geographical information data of the western Sichuan Plateau,and analyzed the characteristics of geological hazards in the highway area.According to the data characteristics and system development requirements,a landslide unit extraction method combining geological survey reports and remote sensing data and a nonlandslide unit selection method based on Gaussian mixture clustering are proposed.Based on the LDA model,a landslide susceptibility evaluation index system was constructed.On this basis,18 landslide influencing factors such as annual average rainfall and distance from rivers were selected,and the relationship between the influencing factors and landslides was studied.A machine learning-based landslide susceptibility modeling framework was established.Completed data preprocessing work such as data cleaning and enhancement,category feature encoding,and data division,and established a feature selection strategy that combines filtering,packaging,and embedding methods.Constructed five landslide susceptibility prediction models of LR,SVM,RF,Cat Boost and LSTM-DNN,evaluated the performance of each model based on the relevant indicators of the confusion matrix and the ROC curve,established the LSTM-DNN coupling model as the recommended model,and realized the impact of the project.Regional landslide susceptibility prediction lays the foundation for geological alternatives selection.A geological alternatives selection model framework based on landslide susceptibility,terrain reconstruction and knowledge reasoning were constructed.An improved two-way ant colony route optimization model is designed,and an alignment corridor scheme with the lowest landslide susceptibility is generated.A road model and terrain surface synthesis algorithm based on constrained Delaunay was studied,which was used to evaluate the influence of alignment schemes on terrain features,and then used to analyze the relationship between alignment schemes and landslide susceptibility,providing information support for alignment scheme comparison and selection.The instance knowledge base and the rule knowledge base are constructed,combined with the spatial relationship reasoning based on the nine-intersection model,the instance knowledge base reasoning based on the K-nearest neighbor algorithm,and the rule reasoning based on the credibility,it is realized that the alignment process can be targeted at specific geological objects knowledge reasoning.Using GDAL/OGR + Scikit-learn/Keras +Civil 3D/Map 3D as the development platform,we have overcome key technologies such as GDAL/OGR-based spatial analysis technology,BIM(CAD)+GIS fusion mechanism,etc.,and developed a highway landslide geological selection The system has designed four functional modules including data preprocessing and analysis and disaster susceptibility prediction,realized the functions of susceptibility prediction and surface synthesis,and verified the rationality and effectiveness of each functional module of the system.