In response to the problem that the downhole environment has a large influence on the geomagnetic data,an HBi-LSTM neural network geomagnetic localization model is proposed.Based on the characteristics of hierarchical LSTM processing different length time sequences and Bi-LSTM fully learning the information of each sequence,the HBi-LSTM model is constructed to collect underground geo-magnetic data by using the built-in magnetometer of mining cell phone to establish a geomagnetic fingerprint database for underground environment,and through HBi-LSTM learning to achieve geomagnetic sequences better corresponding to location tags,after which miners holding mining cell phones collect geomagnetic sequences in random motion to achieve online localization by precisely matching the trained model to the fingerprint database.The experimental results show that the proposed model has better localization performance than the basic LSTM model and can effectively improve the localization accuracy in complex environments.
维普
万方数据