Detection and identification of targets using fusion of HSI and LiDAR data

Abstract

A target in remote sensing may be defined as an object/activity of interest. Detection of targets is an essential requirement in numerous defence and civilian applications. Though primarily, the problem of target detection can be defined as a binary hypothesis testing problem which confirms the presence or absence of targets of interest, however, detection of a target itself may not be sufficient in decision making. Therefore, its classification and identification is also the key. Detection and identification of target using remote sensing data is a challenging task. Different remote sensing data such as hyperspectral (HSI), SAR, LiDAR data etc. have been used for detection of targets. Hyperspectral data, due to its high spectral information, has recently gained momentum in target detection but these may be limited to indicating presence of targets. Therefore, fusion of different datasets may be necessary. In this paper, an approach for detection and identification of two types of targets (vehicle and tree) using decision level fusion of HSI and LiDAR data has been proposed. Firstly, spectral detection using HSI data has been implemented. Next, LiDAR point cloud classification has been performed and used for computing morphological parameters for each target. Lastly the decisions obtained from HSI, LiDAR and morphological analysis have been fused to infer identity of desired targets. Results demonstrated that both types of targets have been successfully identified using the proposed approach. © 2017 Asian Association on Remote Sensing. All rights reserved.