Multi-target Detection And Tracking Of Underground Miners

As one of China's main energy sources,coal occupies an important position in social development.However,due to the complex environment in the mine and the lack of early warning,coal mine safety accidents occur frequently.The coal mine safety problem has never been solved effectively.It is of great significance to implement the strategy of developing coal by science and technology and to use high and new technology to ensure coal mine safety.Establish an intelligent coal mine underground monitoring system to realize the tracking of underground miners and identification of unsafe behavior,and then conduct research on sentiment analysis to achieve intelligent management of coal mines.Intelligent Coal Mine Underground Monitoring System The main research content of this article is to realize the target detection and tracking of multiple miners in the mine.For this aspect of the research,it is necessary to solve the problem of single camera and multi-camera target detection and tracking.The security field and smart city projects also have broad application prospects.Due to the complex background of the mine,low light,and many noises,it is a great challenge to achieve multi-miner tracking.Therefore,research in this direction has significant significance.In order to realize the design of the intelligent monitoring system under the coal mine,through the study of the basic algorithm,this paper puts forward the research plan of multi-target detection and tracking in the mine,the specific contents are as follows:(1)A tracking algorithm based on target detection is proposed.The deep learning method is used to detect the miner's target and extract the miner's feature information to save;second,the Kalman filter algorithm and the Hungarian matching algorithm are used to achieve the target Tracking and matching of IDs of different miners realize the generalization of the network to different environments through transfer learning.Experiments show that this method shows its superiority in dealing with problems such as mine surveillance video and corridor surveillance video under dim light.(2)Multi-camera target tracking.Based on the feature information data ofeach single camera,a multi-camera feature information database is established to convert the target tracking problem into a feature matching problem,and determine the trajectory information of each miner under different cameras through hierarchical clustering,and perform the same target ID under different cameras The merger of the final,through the Hungarian algorithm to achieve the matching of different target IDs.