Moving Detection Research of Background Frame Difference Based on Gaussian Model

At present, there are three algorithms of moving target detection: frame difference, background subtraction and optical flow field. Firstly, those three typical algorithms are analyzed in this paper. In the light of the characteristics of detecting the big moving targets, the frame difference method is improved. Moreover, an adaptive threshold statistics method is put forward to determine the background, which aim is to keep the integrity of the moving target. Background subtraction can accurately determine the moving target. However, the whole image should be computed and its complexity is poor. In this paper, a rapid and precise algorithm of determining the big target is put forward. Firstly, we adopt the improved frame difference method to make sure the full range of moving target. Secondly, the scope which is determined by the frame difference method is considered as the activity´ background. Then detection based on background subtraction is done in a small scope and it can precisely locate the moving target. Then the condition judgment for the automatic update of background is made and the points of background region are calculated by Gaussian model. The condition threshold is set to acquire the fixed background and the moving background, and the fixed background is adaptively updated. Finally, the algorithm is analyzed in detail and the influence of the moving target´ background to the moving target´ precise position is pointed out.