The adaptive visual surveillance system based on the variance of image information

In this thesis, an effective approach for visual surveillance system is proposed. The system capture, analyze and classify the image on the camera automatically, is reduced manpower. This system could determine whether a objects entering, leaving and unchanged base on the variance of the image entropy. First the image preprocessing is employed to remove the noise in the image, then the entropies of gray histogram, horizontal and vertical projection is calculated as image features. Finally, the threshold is found by using the Grey-Prediction algorithm and four conditions are separated. In this project, the amount of memory to store the image is reduced and the condition classification is sample for real-time processing. According to experimental results, the accuracy rate of classifying is about 70%, it would be a useful system to assist the monitoring job.