The Study Of Object Detection Algorithm In Video Surveillance

The moveming object detection is an important content for research in the area of computer vision and video surveillance, target detection in video images have a direct impact on moving target tracking, classification and understanding of senior management behavio processing. This paper summarizes based on the previous motion detection algorithm , studied three aspects,complex environment for motion detection, background modeling and shadow of Gaussian mixture model-based suppression.First of all, this paper analyzes the the frame difference, background subtraction and optical flow detection algorithms, focus on advantages and disadvantages and proposes improvement programs.Second,,It is studied Single Gaussian model and the mixed Gaussian model method two typical background modelling existed, analyzed the contribution of model parameters to the algorithm from multiple perspectives the memory demand and the examination effect aspect. With the systematic study of the detection algorithm of the Gauss modelling, this paper proposed the new thought that organic combination of the sketchy examination's way and the Gauss model's, causes achieves superiorly in the examination effect and the processing speed.At last, On the research of the algorithm shadow suppression, this paper proposed the algorithm shadow suppression based on mixed the Gauss model the shadow to suppress the algorithm. Several steps to the algorithm, first of all, It used the different characteristic between on the HSV the and primitive color space color space to examine the movement prospect whether there was the doubtful shadow, then carry on processing based on Mixes Gaussian Models to inspect the doubtful shadow, completed the task which finally the object examined and shadow suppressed. The experiment proved that the algorithm of shadow suppression based on the Gauss model's movement shadow was able to examine and suppressed the movement goal shadow effectively.