Moving Object Detection in Low-Illuminance Video

Abstract

The security is a main threat exists in every real-time system. The systems like disasterprevention, crime-prevention and boarder security needs highly secured mechanism to alert the people about the incident happening in near about area. Camera system is widely used to monitor the area where security is crucial like for disaster-prevention and crime-prevention. However, at low-illuminance condition, the performance of normal cameras is degrade in great extent. At low-illuminance condition night vision cameras are developed and used but they do not support moving object detection. Also the Gaussian filter technique is used in the system requires the large amount of calculation so the time requirement is also large. Therefore, it is conceivable that the Gaussian filter is not suitable for real time video processing. In this system, input which is low-illuminance video is first denoised using a moving average filter. In that candidate regions of moving object are found by differencing between transformed current image and recent previous image. Moving object is finally decided by combined feature set and motion analysis. Object is tracked by matching object components in ROI. In the motion detection for low-light video images, it is possible to improve the accuracy of recognition by intensity correction and noise removal as preparation. After the detection security system will generate the alert.