Intrinsic Hyperspectral Image Recovery for UAV Strips Stitching

Due to various interferences from natural factors such as changes in the atmosphere and varying illumination conditions, a common issue arises in hyperspectral (HS) imaging, wherein the same material may exhibit different spectra. Unmanned aerial vehicle (UAV) HS imaging is also confronted with this challenge. Moreover, it is essential to merge high-resolution images captured by UAVs because of the limited width of each individual UAV image. This article discusses the aforementioned issues and presents a proficient intrinsic image decomposition (IID) technique for UAV strip stitching to achieve superior quality stitched intrinsic HS UAV images. Initially, an IID energy constraint is introduced for UAV strips to model the IID process of two HS images with a partially overlapping region. Subsequently, an anti-shadow interference mechanism is integrated to mitigate the impact of imaging perspective changes in the overlapping region. However, spectral drift may occur during IID operation in nonoverlapping areas since intrinsic properties recovered from nonoverlapping regions are relatively less reliable. To address this issue, a constraint for spectral drift correction is suggested to rectify the spectral bias existing between overlapping and nonoverlapping regions. The experimental results demonstrate that the proposed method effectively recovers intrinsic properties for UAV HS image strips.