Seismic wave recognition algorithm based on convolution neural network

Abstract

The invention discloses an image recognition and feature extraction method based on a convolution neural network, which mainly comprises the following steps: selecting a sufficient amount of seismic and non-seismic data capable of supporting network training; de-noising and intercepting at the starting point being used to ensure the usability of the data; a convolution neural network being built according to the requirement, and the number of layers and some important parameters being adjusted continuously according to the need and effect in the process of debugging and testing; dropout, BatchNormalization and other methods being used to prevent over-fitting; the seismic data being transmitted to the network in the form of three-component and three-channel for continuous training and debugging, and the network model being tested after many times of training. According to the results, the composition proportion of the training data, the iteration times and the size of the data amount are adjusted. By using this method, we can get 97.17% accuracy of seismic wave recognition.

Description

technical field [0001] The invention belongs to the field of deep learning. Aiming at the problem of distinguishing and identifying natural earthquakes from unnatural earthquakes, a model based on convolutional neural networks is proposed. Through continuous optimization of the network, the model can achieve the effect of accurately distinguishing natural and unnatural earthquakes, and can be widely used in earthquake detection. identify. Background technique [0002] Earthquake is one of the most destructive natural disasters. It has the characteristics of suddenness and destruction, and will produce serious secondary disasters, causing huge losses to people's lives and property. China is a country with frequent earthquakes. Earthquakes mostly occur in five regions including Taiwan Province, Southwest China, Northwest China, North China, and southeast coastal areas, as well as 23 seismic belts. [0003] Natural earthquakes will bring serious personal and property hazards ...

AI Technical Summary

Problems solved by technology

[0012] This algorithm has too high requirements for data volume and precision, and inaccurate data and too small data volume will affect the recognition accuracy of the algorithm.

And limited to the algorithm, there is an upper limit to the recognition accuracy, and the accuracy is not high enough

Images