Species Inspired PSO based Pyramid Match Kernel Model (PMK) for Moving Object Motion Tracking
Anit Chakraborty, Kumar S. Ray, Sayandip Dutta, Siddhartha Bhattacharyya, Anup Kumar Kolya · 2018
The primary contribution of this paper is to deal with different object obscuring one another due to overlap between corresponding support regions. In these circumstances, the competition between two objects elevates to subjugate the overlapping part. In order to effectively design the competition phenomenon, the visual problem needs to be merged with the competition process. We have followed the approach of bag words to extract the textual distribution as feature of the test objects in the video scenes. Furthermore, we have considered discriminative model for tracking the objects with different appearance. Approximating the similarity measures, we have further improved the validation score where we have modelled a linear time matching function which is denoted by Pyramid Match Kernel (PMK) model to synchronize the variable cardinalities to the feature sets. The species inspired PSO framework provides an effective way to track multiple object that are detected and recognized. Our algorithm is evaluated on multiple benchmark datasets with a relative analysis against the numerous state-of-the-art algorithms and outcomes are satisfactory enough to move forward.