Design Of Mine Personnel Positioning System Based On Action Recognition

Coal is the pillar type of energy in the world,China is a coal country,coal production ranks first in the world,but the domestic small and medium-sized coal mine common underground environment is complex,poor conditions and narrow roadway and other security risks,and in-depth management staff can not know the underground work Personnel trajectory and real-time location,in the event of a mine accident caused by trapped personnel,the rescue team will not be able to grasp the location of trapped underground personnel and difficult to work,the rescue work is very unfavorable.Based on the problem of low accuracy and high cost of existing mine personnel positioning system in China,this paper proposes a mine personnel positioning system based on action recognition.The working mechanism of the system is as follows:Underground personnel need to wear data acquisition unit,the data signal is transmitted to the host computer system through the Bluetooth module.The host computer system uses the wavelet packet feature extraction and SVM(Support Vector Machine)classification algorithm to determine the action category(including walking,running,standing).And then according to the X-axis angular velocity of the number of creations to get the number of staff movements,and the running distance of the underground personnel can be determined by calculating the running distance of the underground personnel,and the operation trajectory of the downhole personnel is determined according to whether the person is turning on the basis of the electronic compass.Combined with radio frequency identification(RFID),used as error correction,and finally compared the downhole plan to determine the location of personnel.The specific research content is divided into the following aspects:Firstly,according to the cyclical status of personnel walking and itscharacteristics,the optimal location of the data acquisition unit is obtained.The collected data signal is observed and analyzed,and the wavelet packet feature extraction method is used to process the signal data and extract the eigenvector of the motion signal.Secondly,the key technology in the system is briefly introduced,and the time-frequency analysis of the motion signal is carried out.The method of wavelet packet feature extraction and support vector machine classification is determined,which provides the theoretical basis for the system.Thirdly,the system uses the support vector machine classifier to distinguish the action categories of the personnel.Its classification is based on the characteristic vector of the action signal,because the different actions will produce different vibration energy.The feature vector extracted by wavelet packet feature is sent into SVM for data training and classification,so as to realize the judgment of personnel action type.Finally,a set of mine personnel positioning system is designed,which consists of two parts: data acquisition unit and host computer software system.The data acquisition unit is composed of three-axis gyroscope,electronic compass,Bluetooth communication module,three-axis accelerometer and the minimum system of single-chip microcomputer.The data acquisition and transmission function of the personnel action signal is completed.The software system of the host computer is compiled with the virtual instrument and MATLAB language,to complete the signal data storage,algorithm applications,data processing and downhole personnel positioning and other functions.