Tunnel excavation in karst regions may encounter karst-related geological hazards such as sudden karst water bursts and mud flows.Geological radar is effective in forecasting such karstic and other geological events.However,traditional interpretation of geological radar images heavily relies on expert knowledge,is time-consuming,and is prone to misinterpretation or oversight.This paper explored the use of deep learning technology,specifically designed for end-to-end recognition,in the context of geological radar image object detection and identification.It applied the convolutional neural network algorithm based on Mask R-CNN for intelligent identification of anomalies in karst forecast images generated by geological radar.Under the TensorFlow and Keras frameworks,a training dataset and a test dataset were constructed using data acquired from geological radar.The Mask R-CNN deep learning model was trained on these datasets,ultimately yielding a robust model with better weight parameters for detecting hyperbolic anomalies in karst forecast geological radar images.Experimental results and case studies demonstrate that the Mask R-CNN object detection method achieves excellent performance in detecting and identifying targets within geological radar karst forecast images,significantly enhancing the efficiency of intelligent recognition for geological radar imagery.
NETL
NSTL
维普
万方数据
