Target Detection Using PCA and Stochastic Features

Abstract

Automatic target detection (ATD) system which uses forward-looking infrared (FLIR) consists of two stages: image signal processing and clutter rejection. Images from electro-optical sensors are processed to express target well in signal processing stage. And true targets are well identified in clutter rejection stages. However, it is difficult to process target express well and to identify target from target candidates because they are obscure and there are many target-like objects. We propose new target detection algorithm using PCA and stochastic features. The proposed algorithm consists of two stages; image processing and clutter rejection. Image erosion, dilation and reconstruction is applied to eliminate multiple target candidates that are actually the same, single target and to remove small clutters in image processing stage. Linear Discriminant Analysis (LDA) using principal component analysis (PCA) and stochastic features is applied to clutter rejection. Several FLIR images are used to prove the performance of the proposed algorithm. The experimental results show that the proposed algorithm accurately detects targets with a low false alarm rate.