Human Detection And Tracking Using RGBD Camera

Real-time human detection and tracking is a very important research area and has wide applications in security surveillance, robotics, human-machine interaction, etc. This task is very challenging due to many difficult key points, such as occlusions, complex background, variations in lighting conditions, etc.A new approach to real-time human detection and tracking in cluttered and dynamic environments using RGBD camera is proposed. The notion of Point Ensemble Image (PEI) generated by transforming the original RGBD data to the ground plane is introduced. With this virtual view, human detection and tracking can be significantly simplified.The detector localize the human physique-wise plausible candidates in the first phase. All of the human beings can be effectively retrieved in this phase. The detection results are then refined in a supervised learning and classification second phase. Both RGB and depth information are exploited in this phase, and two novel features, JHCH (Joint Histogram of Color and Height) and HOHD (Histogram of Height Difference), are utilized to characterize the appearance and shape of human body.3D trajectories of humans are finally generated with a basic Kalman Filter model and a JHCH based data association method.The experimental results on the clothing store dataset and office dataset show that the performance of our system outperforms state-of-the-art approaches in terms of both detection accuracy and computational efficiency.