Real time people tracking and counting in visual surveillance

This work presents a real time system to count human beings passed through a doorway. To get the capacity, there were two effective methods. Firstly, using the characteristic of LAB color space, the calculation became simple and people's head was detected easily. Secondly, local features of moving targets were proposed to tracking the human accurately and quickly even the human turned around. With this algorithm, the system counted the passed human accurately without detecting the target in every frame. This system produces good tracking and counting results at a frame rate of 15 similar to fps for image size of 320x240, so it can be very useful in some real time visual surveillance.