Development of Adaptive Tracking using Advance Filter and Selection Features Method

Abstract

Recently, Kalman filter(KF)-based algorithms of tracking had demonstrated to be effective, however, their efficiency is limited by fixed feature selections and the possibility of model drift. In the presented research, we offer a new adaptive feature selection-based tracking approach that maintains the KF’s excellent discriminating power. Depending on scores of confidence regarding features in every one of frames, the suggested approach might select (automatically)either SIFT feature or the colour feature for the tracking. With a use of KF, a response map related to the SIFT features and color features are retrieved first. The color features that distinguish the luminance from the color are extracted using the Lab color space. Second, the average peak-to-correlation energy is used for the determination of the confidence region and the target's possible location. Finally, a total of 3 criteria have been utilized in order to choose the appropriate feature for present frame in order to execute adaptive tracking. On OTB benchmark datasets, the experimental findings show that the suggested tracker performs better in comparison with other state-of-art techniques.