Surface Vessel Tracking in Airborne Infrared Imagery

---

Tezin Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Mühendislik Fakültesi, Elektrik ve Elektronik Mühendisliği Bölümü, Türkiye

Tezin Onay Tarihi: 2019

Öğrenci: AHMET ÇAKIROĞLU

Danışman: İLKAY ULUSOY

Özet:

Target tracking can be defined as continuously locating the object of interest in consequent images. Tracking surface vessels in infrared imagery is an exceptionally challenging case of visual target tracking. In a typical scenario both the target and imaging platform exhibit manoeuvring movement, causing the appearance of the target to change rapidly and significantly during the course of tracking. Furthermore there are cases where target actively attempts to avoid being tracked by firing hot flares to confuse the tracker or block the view of the tracker. In some cases target also cools itself down to background temperatures with special equipment to blend itself with the background and avoid being seen. In this thesis one of the popular general object tracking algorithms is improved by transfer learning and developing an occlusion detection mechanism. Discrimination power of the tracker is increased by transfer learning and occlusion detection capabilities enabled the tracker to reacquire the target after occlusion. Performance of proposed algorithm and other several distinguished target tracking algorithms are compared on our infrared surface vessel image dataset. Image dataset consists of synthetic images acquired during challenging naval combat scenarios and categorized by their respective challenges such as confusion, low intensity and occlusion. It was seen that the proposed algorithm had superior performance to other tested algorithms.