Study on Engineering Geology Petrofabric Regionalization of Slope Surface Based on Multi-source Data

Aiming at the problem of geotechnical engineering property regionalization of slope area which is concerned in the prevention and control of geological disasters, this paper proposes a method of geotechnical engineering property regionalization of slope shallow layer based on multi-source data fusion. Taking Zigui to Badong section of the Three Gorges Reservoir as the study area, the stratum lithology, slope structure, slope, remote sensing image spectrum and texture information are selected as the regionalization factors. Machine learning samples are obtained through data collection, remote sensing interpretation and field investigation. After learning the samples with support vector machine algorithm, the geotechnical engineering properties of the study area are divided into three types: hard rocks, soft and hard interbedded rocks, soft rocks (including soil). Through the confusion matrix accuracy evaluation, the overall classification accuracy is 74.6%. Indicating that the regionalization results are relatively accurate and reliable. The results show that, with the participation of remote sensing image spectrum and texture information, the area with thick soil on the slope can be well identified divided into soft rocks type.