Abstract: Basic geological phenomena in a hydraulic cavern can be a reflection of the occurrence of geological disaster. An efficient classification and identification of such information are crucial to understanding the distribution of structural planes in the cavern and the properties of its surrounding rocks to guide its further exploration. This task usually relies on manual work, but such traditional methods are time-consuming and labor-intensive at a low level of automated analysis. In this work, we use a variety of deep learning models and machine learning models to analyze the images of basic geological observations of hydraulic cavern foundation; and the cavern phenomena are classified through applying different deep learning models and the methods of Softmax classifier, random forest, and support vector machine. By comparing and selecting models with better performance for coupling, we developed a satisfactory image recognition model for the cavern geological phenomena, achieving automatic identification and analysis of the caverns, so that the workload of geological engineers is reduced significantly.

Key words: hydraulic cavern, geological image, deep learning, machine learning, intelligent classification