Parameters affecting target detection in VNIR and SWIR range

Abstract

Hyperspectral data due to its fine spectral resolution facilitates detection of targets with high accuracies and thus has been used in many applications. A target detection application differs from most other remote sensing applications in the fact that targets are often too small and contained in only a few pixels. However, HSI detection gets limited by several parameters which can be broadly categorized in three groups: sensor parameters (noise, calibration etc.), spatial parameters (size, shape etc.), scene parameters (illumination variation, background etc.). These parameters may have varying effect on detectability of targets and thus need further investigations. Two specific spectral ranges VNIR and SWIR have been selected based on literature review to study effect of four scene parameters namely illumination, background, material composition and color of target on detection of known targets. The detection has been implemented using three popular detection algorithms namely, ACE, MF, SAM. Results indicate that scene parameters play a major role and affect detectability of targets differently in different spectral ranges.