Agriculture Multispectral Uav Image Registration Using Salient Features and Mutual Information

Abstract

Multimodal image registration has been studied for a long time as a necessary pre-processing step to extract relevant information from the studied images. In this direction, agriculture remote sensing has evolved to use multispectral sensors and faces challenges since the application of classic solutions is not suitable. This paper preliminarily explores the benefits of applying Mutual Information (MI) based on SIFT points for image registration to agriculture remote sensing multi-spectral evaluated on a self-developed public database of images through a fixed-wing Unmanned Aerial Vehicle (UAV) equipped with a multispectral sensor operating within parameters that would apply to crops inspection in real life. Our preliminary results have shown a marginal improvement of MI after registration highlighting that we may apply it to improve the registration of agriculture remotely sensed images. This small variation of MI shows that there is room for improvement.