In this paper, an approach for tracking objects using motion information over compressed video is proposed. The input data is a set of blobs or regions obtained from the segmentation of H264/AVC motion vectors. The tracking algorithm establishes correspondences between blobs in different frames. The blobs belonging to the same object must satisfy a set of constraints between them: continuity, temporal coherence, and similarity in attributes like position and velocity. The uncertainty and dispersion inherent in motion data available from compressed video makes fuzzy logic a suitable technique to achieve good results. Then, the blobs are represented as sets of linguistic data called linguistic blobs and all the operations are performed over this structure. A major contribution of this work is the design of a tracking process that is able to operate in real-time because the size of input data is very small and the computational cost of operations is low. The main calculation cost is reduced by using simple mathematical computations over the linguistic variables. © Springer-Verlag Berlin Heidelberg 2014.
CITATION STYLE
Rodriguez-Benitez, L., Moreno-Garcia, J., Giralt, J., del Castillo, E., & Jimenez, L. (2014). An Application of Soft Computing Techniques to Track Moving Objects in Video Sequences. In Advances in Intelligent Systems and Computing (Vol. 214, pp. 291–303). Springer Verlag. https://doi.org/10.1007/978-3-642-37832-4_27
Mendeley helps you to discover research relevant for your work.