Detection and Tracking of Platforms above Sea-Level in Infrared and Visual Band Videos using Bag-of-Features Technique with Scale Invariant Feature Transform

In this study, platforms above sea-level are detected and tracked using Bag-of-Features technique (BoF) with scale-invariant fea- ture transform (SIFT) in infrared (IR) and visual (VIS) band videos. Features in the training set corresponding to platforms above sea-level and background are firstly clustered in offline mode and these features are used for target detection using BoF technique. With these features, targets are matched in the subse- quent frames and the matchings are pruned with a set of heuris- tic rules. Target tracking performance is compared with respect to the ground-truth target positions for different IR and VIS band videos obtained in field trials and with an optic-flow based method. The results show that platforms above sea-level can be detected and tracked using SIFT features and BoF technique with high accuracy. 1. Giris¸