Alignment of UAV-hyperspectral bands using keypoint descriptors in a spectrally complex environment

Abstract

ABSTRACTNear earth imaging spectroscopy has gained popularity in the recent past due to its increasing capabilities to acquire data with unprecedented abundance of spatial resolution and spectral purity. Unmanned Aerial Vehicle (UAV) based portable hyperspectral sensors employing snapshot based scanning mechanisms further advances these capabilities with improved spatial co-registration of pixels. However, unavoidable motion of UAV pose challenges in spectral co-registration which is pronounced in spectrally complex environments of heterogeneous ecosystems. In this study, the different feature descriptor techniques such as Harris-Stephens Features (HSF), Min Eigen Features (MEF), Scale Invariant Feature Transformation (SIFT), Speeded-Up Robust Features (SURF), Binary Robust Invariant Scalable Keypoints (BRISK), and Features from Accelerated Segment Test (FAST) were evaluated to align hyperspectral bands in a spectrally complex environment. A band alignment workflow was devised and operated in different band-wise arrangements (spectral order and temporal order) of the acquired hypercube. The co-registration accuracy between the adjacent band pairs (reference-transformed) upon registration was estimated using Root-Mean-Square Error (RMSE) and Pearson’s Correlation Coefficient (PCC) based approaches. Furthermore, a standard transformation workflow was used to evaluate the efficiency of the different feature descriptor based band-to-band registration approaches.