Real-time illumination feedback system for adaptive background subtraction working in traffic video monitoring

In this paper, we presents a real-time adaptive background subtraction algorithm working in traffic video monitoring to solve a classic problem of background subtraction “sudden illuminance changing”. We design an illumination feedback kinematics to update a background image. It consists of two main phases, illumination approximation for feedback and background image adjusting. In the illumination approximation, illumination difference between a current image and the background image is calculated. Then, two weight parameters for the background image adjusting are introduced to improve performance of background subtraction. The real background pixels from background subtraction are updated to a new background image for next iteration, in the background image adjusting phases. The system achieved a high performance with maximum 0.959% false negative detecting (foreground pixels are justified as background pixels), and high effectiveness with 121.3 frames per second (CPU Intel i5). The experimental results show the efficiency and effectiveness of our algorithm working in traffic video monitoring in real-time condition.