HUMAN TRACKING USING SURF AND PARTICLE FILTER

Hitman tracking is a vital research topic nowadays, due its broad applicability and importance. Many algorithms were developed for tracking process including particle filter, which is used to solve non-linear and non-Gaussian problems efficiently. However, particle filter suffers from degeneracy problem and considerable computational time. In our proposed framework, we worked on enhancing particle algorithm performance by integrating SURF features with particle filter. To the best of our knowledge, it is the first time to integrate SURF features into particle filter where extracted features are used partially to feed the particle filter with samples/particles. The added value here comes from the fact that SURF is one of the most fast descriptors which generates a set of interesting points which are invariant to various image deformations ( scaling, rotation, illumination) and robust against occlusion conditions during tracking . Hence, particles used to track human will not be chosen randomly as done in standard particle algorithms, instead they are chosen with the help of SURF. The experimental results, performed using KTH action database, proved enhancements in solving degeneracy problem, reducing computational costs and well performance under lightning, scaling, indoors and outdoors conditions