Abstract: Aiming at the problem of moving object detection based on mixture of Gauss background model，mainly the false detection of light mutation and the“ghosting”of sudden moving object，an algorithm of moving object detection of mixed Gauss background model based on three frame differencing is designed． According to the proportion of the foreground of the image to judge whether the light is changed，the three frame difference method is used to divide the background area，the moving area and the exposed background area． According to the illumination，the learning rate of each region can be changed in time to adjust the update speed of the background of the mixed Gauss model，and a new method based on three frame difference and adaptive learning rate is proposed to update the background of mixed Gauss model． This method makes the new background model produced by abrupt illumination and sudden motion of the target rapidly updated，improving the detection result of moving objects in these two cases． The experimental results show that the moving object detection algorithm can avoid large area error phenomenon of illumination mutation，and solve the “ghost”problem of moving object．

Key words: moving object detection, light mutation, three frame difference method, mixed Gauss model, background updating