Classification and extent determination of rock slope using deep learning

The classification of rock slopes and the determination of rock extent are essential for slope stability analysis. At present, the efficiency of artificial methods is low and affected by subjective factors. A convolution neural network model for rock slope image analysis is established based on the Tensorflow. 80,000 rock slope images are extracted and compressed using convolution and pooling. Then a network model is trained to automatically identify and classify rock slopes. Using the training set and the rock slope image in the test set to test and analyze the model, the training set accuracy rate is 98%, and the test set accuracy rate is 90%, which indicates that the network model after training is robust. Based on the color of different rocks in the slope, the extent of different kinds of rocks in the rock slope is determined using a deep learning regression operation. In order to verify the effect of the algorithm, a standard color rock slope image is selected for a simulation experiment, and the boundary detection is accurate. Finally, the deep learning network model is used to quickly and automatically identify and classify rock slopes. The rock slope information obtained by image identification is imported into independently developed GeoSMA-3D software, which is an important parameter for determining the grade of rock slopes.