Research On The Method Of Shallow Underground Source Scanning Location Based On Deep Learning

The shallow underground source location technology is a hot spot in the field of underground space location research.This technology can effectively solve the problems such as the location of the underground blast point of high-value ammunition and the measurement of penetration trajectory,and it is also an important means to realize the civil problems such as geological monitoring,engineering blasting,cultural relics anti-theft monitoring,and coal mine investigation.Because the underground velocity model is complex,which cannot meet the high precision reconstruction of the underground energy field,so the energy focusing region in the generated energy field image is fuzzy and the resolution is low.As a result,it is difficult to locate the shallow underground source and its positioning accuracy is low.In order to solve the above problems,this paper has carried out a research on the method of shallow underground source scanning location based on deep learning.Firstly,using the theory of wave field reciprocity,the 3D energy field image sequence is reconstructed as the input sample by grouping cross-correlation imaging conditions.Secondly,from the perspective of energy focus recognition,the energy focus identification and location method based on convolutional neural network is proposed.First,an attentional mechanism based 3D-Densenet(dense connection network)was established to extract the features of the energy focus region.Then,the 3D spatial pyramid pooling model is constructed to extract the multi-scale features of the above-mentioned energy focus area,and finally the identification and location of the energy focus(source)are realized.Thirdly,from the perspective of energy focus search,the energy focus search and location method based on reinforcement learning is proposed.The source location process is regarded as a Markov decision process to build the location model.The 3D energy field is regarded as the environment in the model.After the search box(agent)makes the decision through the deep decision network,it will get positive or negative evaluation according to the performance.The feedback results will make it gradually approach the real source point,so as to realize the source location.Finally,the field test verifies that the energy focus search and location method based on reinforcement learning is better than the energy focus identification and location method based on convolutional neural network,and its positioning accuracy reaches 92%.The experimental results show that the energy focus search and location method based on reinforcement learning can effectively improve the positioning accuracy of shallow underground sources.