A Moving Foreground Expansion Method Based on the Gaussian Distribution
Yanhua Li, Wei Li, Qi Xiang · 2013
With the development of computer science, intelligent video surveillance technology has been widely used, moving target detection becomes an important part in the field of intelligent video surveillance. Moving foreground extraction, which is the first step of moving target detection, decides the accuracy of moving target detection. Traditional frame difference, background subtraction and other moving foreground extraction algorithms do get the extracted results but still have some shortcomings, such as disabilities and holes. So a moving foreground expansion method based on Gaussian distribution is proposed in this paper. This method utilizes the theory of Gaussian distribution to establish the Gaussian kernel on the boundary of the moving foreground. Next, the mean and variance of the Gaussian kernel is calculated. And then the necessary probability can be obtained with the mean and variance. At last, we can determine whether to expand according to the contrast result of the probability and the stated threshold, making the missing parts get an effective supplement and expansion. The experiments show that the method can detect moving targets more accurately due to effectively complementing the moving foreground.