Research On Detection And Tracking Of Video Targets In Coal Mines

With the rapid development of new-generation information technology in artifcial intelligence,cloud computing,and big data,informatization has become the development trend of coal mining enterprises.Target detection and target tracking of underground monitoring video in coal mines,as a means of analyzing and understanding coal mine's target behavior,have become a research hotspot in coal safety production.The study of target detection and tracking methods in specific environments in coal mines can provide new technical support and an effective guarantee for coal mine safety production.This paper first impoves the preprocessing algorithm of coal mine's video images,and then studies the traditional Gaussian mixture model(GMM)target detection algorithm based on the improved preprocessing algorithm to make it suitable for the underground environment of coal mines,and finally adopts the improved CamShift target tracking algorithm to track the moving target in the coal mine.The research content has the following three aspects:This paper first improves the preprocessing algorithm of underground video images,and then studies the traditional GMM target detection algorithm based on the improved preprocessing algorithm to make it suitable for the underground environment of coal mines,and finally adopts the impro ved CamShift target Tracking algorithm to track the moving target in the coal mine.The research content has the following three aspects:(1)Preprocessing of coal mine video images.Dim lights,coal dust,water mist and transmission equipment in underground coal mines have led to low contrast and high noise in underground video images.In order to solve this problem,wavelet transform combined with an improved dark channel fog removal algorithm is used to preprocess the underground video image.The improved algorithm improves the contrast of the coal mine image,and achieves good denoising and fog removing effect while preserving the detailed information of the image.(2)Coal mine video target detection.When performing target detection in underground coal mines,there are still ignoring the correlation between pixels,resulting in low algorithm calculation efficiency,inefficient detection algorithms,"holes" in the detection result images,inaccurate detection due to sudden changes in illumination,and fixed model correlation thresholds.Inaccurate model update,slow speed and other issues.For this reason,this paper proposes an improved GMM target detection algorithm.First,an improved block modeling algorithm is used for the video,and then the adaptive learning rate of the GMM is designed based on the proportion of the foreground in the background image,and the three-frame difference method is combined to suppress the impact of illumination on the detection.The improved algorithm improves the speed and accuracy of Gaussian modeling,and suppresses the influence of illumination changes on coal mine's target detection.(3)For the tracking of underground moving targets,an improved CamShift algorithm is proposed.By reducing the color value of the background color of the search area,increasing the color difference between the target and the background,and suppressing the interference of the similar color background on the target tracking,the Kalman filter algorithm combined with the CamShift algorithm is used to predict and track the position of the moving target in the coal mine's video.This method can effectively overcome the influence of the similar colors of the foreground target and the background on target tracking,and improve the accuracy of coal mine's target tracking.