Application of Convolutional Neural Network to GIS and Physics

Human life and property are often seriously threatened and lost due to natural disasters such as earthquakes. As a spatial information system, the geographic information system (GIS) can collect, store, and manage the local or whole related physical data of the surface space to be measured with the support of software and hardware. The physical data is collected through GIS for performance testing. The data are collected from the aftershock event records of the Wenchuan earthquake. Among them, 14,000 Wenchuan earthquake events are used as the original data set, and 8,800 aftershock events are used as the test data set. Seismic data involves the detection of multiple physical quantities, which makes the seismic data gradually increase, many data have no obvious linear relationship, and traditional detection methods are difficult to meet the detection requirements. The artificial intelligence method led by a convolutional neural network (CNN) can perform pattern matching on complex nonlinear variables, and models with general characteristics can be generated from different seismic waveforms for the prediction of seismic waveforms. The results show that GIS can effectively intercept and collect seismic physical signals. The training and detection accuracy of CNN combined with GIS physical data is higher than 90%. Compared with traditional training methods, CNN is obviously superior in detection accuracy and recall rate. At the same time, a large number of microseismic events that are easily missed by manual selection can also be found.