A Method for Multispectral Images Alignment at Different Heights on the Crop

Multispectral imaging (MSI) for agricultural applications is playing a key role in plant stress assessment. However, one of the main problems is the misalignment of spectral bands provided by these instruments. The approach proposed in this study considers various distances from the target (500 mm to 1500 mm with a step size of 100 mm) and applies corrective shifts to achieve accurate registration among bands. Through a comparative evaluation of two alignment methods, Checkerboard (CB) and Discrete Fourier Transform (FT), this research aims to provide an effective solution for accurate image registration by facilitating reliable spectral analysis. Specifically, the proposed method involved the analysis of alignment-related offsets among the tested methods. In addition, the study explored the extraction of vegetation spectral indices for vegetation analysis and discrimination between healthy and diseased plants and evaluated their relationship with the quality of alignment obtained at different heights. The results confirmed the trends in the changes in offsets as the target distance varies, showing satisfactory accuracy in the alignment of raw spectral images at different distances, with an error of about 1 pixel. Among the vegetation indices used, the Normalized Difference Vegetation Index (NDVI) proved to be capable of discriminating between healthy and nonhealthy leaves. The study aims to establish a framework applicable to remote sensing and agricultural monitoring, providing a valuable tool for monitoring plant health.