Research On Vision-based Object Detection And Tracking Algorithm In Underground Coal Mine

There are many problems such as uneven illumination,low overall illumination,and more coal ash and dust in the coal mine,which bring great challenges to the application of target detection and tracking technology in underground coal mine.Therefore,how to detect and track objects accurately in coal mine surveillance video is an important research topic of intelligent coal mine video surveillance system.Based on these,the main contents of this thesis are as follows:(1)In view of the problems of large noise,insufficient illumination and low contrast in the monitoring video of coal mine,the method of combining wavelet denoising with adaptive histogram equalization is proposed in this thesis.This method is mainly used to denoise and enhance the video image.The result is that the noise is effectively eliminated,and the brightness and contrast of the image have also been improved to a certain extent.(2)The influence of light change and light spot interference on the underground environment leads to the inaccuracy of the target detection results.In this thesis,a target detection method which combines the double threshold background difference method and the inter frame difference method is proposed.In this method,First,the interframe difference method is used to determine the region of interest in the video image,and then the region and the background model are calculated by the difference operation.Finally,according to the characteristics of the light spot,the dual threshold two values and the mathematical morphology are used to deal with the obtained differential images,so as to further determine the target area accurately.The experiment shows that the target detection algorithm proposed in this thesis can eliminate the interference of the light spot of the mine lamp to some extent,and also has good robustness to the illumination change.(3)Aiming at the problems of low discrimination and interference of downhole targets,in this thesis,a feature extraction algorithm based on adaptive PCANet is proposed in this thesis.The algorithm introduces the cumulative contribution rate to determine the number of the main filters in the PCA layer,and simplifies the PCANet tuning process.Combined with this algorithm,the target tracking algorithm of mine monitoring video presented in this thesis has better performance.Firstly,the tracking algorithm uses particle filter to obtain the region of interest in the video image.Then the proposed adaptive PCANet is used to extract the depth feature of the selected region,and the target area is determined by the SVM classifier.The experiment shows that the target tracking algorithm in this thesis shows good robustness in the complex coal mine environment,and can also effectively track the target under the circumstances of light interference,occlusion,deformation and so on.