Robust object tracking in the X-Z domain
Ankita Sikdar, Yuan F. Zheng, Dong Xuan · 2016
The detection and tracking of objects such as humans has primarily been carried out using RGB cameras with additional sensors such as infrared sensors, ultrasonic sensors, etc. and therefore two-dimensional tracking (x-y tracking) has been widely researched. With the advent of depth cameras, depth based tracking has gained prominence as one can exploit the third dimension (z). While prior work on depth based tracking using Kinect sensors focuses mostly on depth based extraction of objects to aid in tracking, this paper introduces the idea of tracking human objects in the x-z domain instead of the x-y domain. Tracking is done by particle filters which are propagated based on the motion model in the horizontal-depth movement framework. A joint color and depth histogram model, used to represent a human being, is adapted based on the occlusion status of the object, and the combination with position and size with respect to the tracked human object in the previous frame is used to filter the depth segmented objects in each frame. Particles, depicted by patches extracted in the x-z domain, are associated to these observations based on a closest match according to a likelihood model and then a majority voting is employed to select a final observation, based on which, particles are reweighted and a final estimation is made. The addition of the depth dimension in motion propagation and tracking alleviates challenges faced due to change of object appearance, illumination changes and partial occlusion (or full occlusion in some cases). Results show that a tracker based on x-z propagation outperforms a similarly designed baseline x-y tracker.