Heuristic algorithms for fast and accurate tracking of moving objects in unrestricted environments

Abstract

Tracking of objects is a basic process in computer vision. This process can be formulated as exploration problems and thus can be expressed as a search into a states space based representation approach. However, these problems are hard of solving because they involve search through a high dimensional space corresponding to the possible motion and deformation of the object. In this paper, we propose a heuristic algorithm that combines three features in order to compute motion efficiently: (1) a quality of function match, (2) Kullback-Leibler measure as heuristic to guide the search process and (3) incorporation of target dynamics for computing promising search alternatives. Once target 2D motion has been calculated, the result value of the quality of function match computed is used in other heuristic algorithm with the purpose of verifying template updates. Also, a short-term memory subsystem is included with the purpose of recovering previous views of the target object. The paper includes experimental evaluations with video streams that illustrate the efficiency for real-time vision based-tasks.