A practical algorithm for exception event detection for the home video security surveillance

This paper introduces a new method of moving object detection by combing pixel illumination with its chroma in YUV color space. In a practical application, the surveillance scenes are possibly complex and there may be disturbing background objects, such as swaying leaves or waving curtains, which can bring about false alarms. To eliminate these disturbing objects, we give a new algorithm for background maintenance with 3 key values. It is very effective and robust to a complex scene. In order to reduce false alarms, a real-time and practical algorithm for exception event classification is also introduced