Abstract: Detecting and tracking moving objects in the dangerous area of underground mine plays an important part in coal mine monitoring system, which can issue early warning and prevent dangerous accidents in combination with an interlocking system.For the objects moving in environment with slowly changing or suddenly changing illumination, it is proposed to realize the detection by an improved on-line adaptive learning rate Gaussian mixture model in combination with frame difference, and the background modeling method in combination with color gradient and mathematical morphology, and to remove the shadow by HSV color model, thus enabling a rapid and accurate location of moving object.

Key words: Mine safety, Gaussian mixture model, Shadow suppression, Monitoring system