Motion characteristics estimation of animals in video surveillance

This paper proposes a system for the motion characteristics estimation of multiple objects with uncertain quantity. The system employs background subtraction method based on Gaussian Mixture Model to detect objects and tracks them through an improved algorithm combining Camshift with Kalman filtering. By analyzing the directions and trails of targets, the system can figure out their pixel distance and velocity. Experiment conducted realizes the proposed system, verifying its performance on motion characteristics estimation.