A visual method for real-time detecting persons in complex scenes

This paper presents a method for detecting persons in complex scenes which are called backgrounds. A panoramic image about the background is created before the person come into the background. When the persons come in front of the background, a detecting process is started up. We propose an algorithm for determining the camera motion parameters, which is used to get the background image hidden by the persons from the panoramic mage. The current captured image and the background image are compared to detect the persons using the background subtraction algorithm based on logarithmic intensities. Experiments show that the proposed method can be real-time run on a high-performance personal computer.