Uzun Dalga Kızılötesi Hiperspektral Görüntülerde Hedef Tespiti

Abstract

In this thesis, target detection from hyperspectral images has been investigated using anomaly detection and endmember extraction. The aim of anomaly detection is to detect the deviations from the normal background behavior without any prior information about the data. In hyperspectral image analysis, Reed-Xiaoli (RX) algorithm is a commonly used anomaly detector. It first models the background as a multidimensional Gaussian distribution and then computes how much a test vector deviates from the background model. In this thesis, six different RX-based anomaly detectors, namely the global RX, local RX, dual window RX, subspace RX, kernel RX and the global RX combined with a uniform target detector have been tested on long-wave infrared (LWIR) hyperspectral images. Several factors such as parameter selection, resilience to noise, computational complexity, effect of window size have been examined and the detection performances have been compared based on ROC curves.