Tracking Vehicle and Faces: Towards Socialistic Assessment of Human Behaviour

The modern day robots can do a variety of tasks with great efficiency, however their utility is limited due to the non-social behaviour of the robots. For the same it is important to assess the human behaviour in diverse conditions to as to eventually make robots socialistic in nature. Object & people tracking is an excellent field of computer vision in which we have tried to detect and track multiple vehicles and people at outdoor traffic environment and indoor office environment respectively. A background subtraction algorithm is applied for vehicle detection. Kalman filter is used to predict the estimated position of every vehicle in the next frame and updating of new track. Some vehicles are also detected in cluttered scenes. Every moving vehicle is counted in video frames. Multiple face detection and tracking is an also attractive field of computer vision. The faces are behaviourally very different to vehicles and indoor scenarios are also very different to outdoor scenarios. Hence a different methodology to track people is used. In this paper, we have tried to detect and track face of multiple people on two different datasets with different height of camera. Point feature is extracted and compared it in the successive frames to track face of multiple persons. Every face is bounded by rectangular shape with unique identity. We have also counted total number of faces in the frame sequences. Results on both vehicle and face datasets show promising results and the proposed methodology can accurately track the trajectories. The output of the research is a good dataset to assess human behaviour to be used in social robotic applications of the future.