Identification of military vehicles in hyper spectral imagery through spatio-spectral filtering

The application of Hyper Spectral Imagery (HSI) for identification, classification and status of a specific material based on its spectral characteristics has been demonstrated by the researchers in past. In recent years, the use of Hyper spectral imagery in areas relating to tactical detection and classification of military vehicles is growing interest. However, literature on the suitable algorithms or methods for these type of applications are scarce if non existent. In this paper authors are proposing a method for detecting sub pixel sized military vehicles in acquired hyper spectral imagery. In the proposed approach Region Of Interests (ROIs) identified with Reed-Xiaoli (RX) anomaly filter, are processed using spectral-spatial information for identifying military vehicles. Performance of proposed method is analysed on Hyper Spectral Image (HSI) data set constructed by embedding two types of military vehicle signatures in HSI data cube at random locations. Principal Component Analysis, Anomaly detection (RX) and Spectral Angle Mapper (SAM) classification algorithm are applied to the data set being analysed. This work shows that using proposed method detection and discrimination of military vehicles is feasible with high probability of detection and low probability of false alarm.