People tracking using robust motion detection and estimation

Real world computer vision systems highly depend on reliable, robust retrieval of motion cues to make accurate decisions about their surroundings. In this paper, we present a simple, yet high performance low-level filter for motion tracking in digitized video signals. The algorithm is based on constant characteristics of a common, 2-frame interlaced video signal, yet results presented in this paper show its applicability to highly compressed, noisy image sequences as well. In general, our approach uses a computationally low-cost solution to define the area of interest for tracking of multiple, moving objects. Despite its simplicity, it compares very well to existing approaches due to its robustness towards environmental changes. To demonstrate this, we present results of processing a sequence of JPEG-compressed monocular images of a parking lot in order to track pedestrians, cars and bicycles. Despite a high level of noise and changing lighting conditions, the algorithm successfully segments a moving object and tracks its position along a trajectory.