People detection and tracking via 3D range sensor

With the rapid development of range sensing devices, nowadays 3D range sensor is able to collect high-quality range image sequences and plays an important role for crowd surveillance applications, especially in congested situations. In this research, we propose a novel framework for detection and tracking of people in the surveillance area by using the Kinect sensor. Sensors are set at the height of 3-5m and designed to scan the objects below it. People detection is performed on the 3D point set based on an improved version of mean-shift clustering algorithm, where additional weight is considered and the weight of the local peak is further enhanced so that it can locate people's heads at different heights more accurately. Then tracking analysis is performed by utilizing the spatial-temporal consistency. Experimental results demonstrate the effectiveness of our proposed method.EI