Effective use of dense optical flow for path recognition

Separation of traversal path from the non-traversal path area is a fundamental problem in computer vision and an active area of research in the intelligence transportation systems community. In spite of many years of research, general purpose traversal path-detection mechanism is still a very challenging task because of the dynamic nature of the variables involved and the ever-changing geographic terrains. Among the various methodologies addressing the issues of traversal path detection, schemes that employ path detection are good but lack robustness whereas other procedures such as template matching are static in nature and one may even consider a system involving only optical flow which may be robust in comparisons to others but cannot be applied to various region of interest and is thus unable to solve the above mentioned problem as a single entity. In this paper, we present a methodology wherein we apply optical flow in synchronization with dynamic template matching to generate the traversal path area. The problem is partitioned into several categories; each category is processed according to a particular method and the output may serve as the input for the other classes. Information is also provided on real-time performance of the algorithm and the results are illustrated to describe the efficiency and robustness of this approach. The algorithm having tested over various environmental conditions is found to be effective and robust.