Research On Identification Of Miners' Violation Behavior Based On Kinect

The safe production of coal mines is one of the important factors for the sustainable development of coal industry and the downhole video surveillance is an important part of coal mine safety production.Since most of the coal mine safety accidents are caused by the illegal operations of underground workers,so video survillance of the illegal operations of underground workers is very neccessary.At present,underground video survillance is usually based on traditional cameras,there will be some monitoring blind areas due to no monitoring in certain areas of low illumination.This paper obtains coordinates of human skeletal points by using the Kinect sensory sensor,and an improved dynamic time warping algorithm is proposed to establish a real-time miners' violation behavior recognition system.The main work of this paper is as follows:(1)The DTW template clustering algorithm is improved,which effectively solves the problem that the DTW algorithm is slow in template training in the violation behavior recognition.This method directly divides the training set into a suitable subset and provides the appropriate initial template loss,which avoids the gradual decomposition of subset and template vector by traditional clustering algorithm,gradually increasing the time consumed.The exprimental result shows that the improved algorithm effectively sloves the problem of slow training of identifying templates for irregularities.(2)As for two typical problems in the DTW algorithm—singularity problem and time complexity,this paper proposes two types of improved algorithm.First,a piecewise linear approximation combined with an adaptive weight dynamic time warping algorithm is proposed to solve the singular point problem(abbreviated as PLA-SWDTW).Then,the algorithm of using dynamic global programming in the global matching path is proposed to solve the time complexity problem.We carry out ecperiments and simulations on the two improved algorithm,The average recognition rate of the PLA-SWDTW algorithm in the SDUFall Dataset data set reached 95.33%,the dynamic global programming algorithm reduced the average recognition time by 46.47% in the simulation experiment.(3)Based on the improved algorithm,a mine personnel identification system is established,which was carried out in the laboratory and in the mine.The experimental results show that the average recognition rate of the improved algorithm is 95.5%,and the average recognition time is reduced by 44%.Therefore,the system can effectively identify the violation behavior of mine personnel in the real scene.