Research On The Motion Detection Algorithm For The Video Surveillance Of Coal-bed Methane

With the rapid and sustainable development of economy, the existing energy has been gradually unable to meet the development demands. As a result, the conflict between energy supply and demand is becoming increasingly prominent. As a new kind of clean energy, further exploitation for the coal-bed methane(CBM) has attracted a great deal of attention of the government. However, by considering that the CBM field is usually located in remote mountain areas and the mines are decentralized, the safety inspection is inefficient only rely on the staff. Therefore, the CBM video surveillance is urgent to promote safety production timely and effectively. When there is an exception in the monitor scene, the video monitoring system can warn timely. As the basis of video surveillance, the motion detection is very important.The algorithm of motion detection has been one research focus for decades. By doing plenty of further researches, many effective algorithms of motion detection have been found by many researchers at home and abroad. But when refer to the complex monitoring environment, there are still some difficulties in detecting the moving targets accurately and timely.In this paper, several common motion detection algorithms are introduced firstly. Gaussian mixture background modeling is studied and the simulation result is analyzed. The analysis found that it is useful when the scene is static or simple dynamic but not ideal for the complex monitoring scene. What’s more, this algorithm is based on the prior knowledge of the background which is difficult to be obtained. As a result, the practical application of the algorithm is restricted by the condition. Next, nonparametric kernel density estimation is analyzed according to the environment of CBM mining. Gaussian kernel density estimation background model is founded, and then the object detection, the noise reduction and the background update are realized. In addition,the simulation result is better using the bandwidth presented in this paper. The algorithm is confirmed more effective through simulation experiment based on indoor and outdoor case. Compared to the previous simulation, the detection result is clearer, the connectivity is better and the speed is faster. What’s more, the algorithm can suppress noise more effectively than the mixture Gaussian background model. All these make the nonparametric kernel density estimation satisfy the practical requirements fully.