Automatic fine co-registration of UAS multispectral images: a case study on a vineyard in Tuscany, Italy

This work presents an automated preprocessing patch for the fine co-registration of multispectral (MS) images captured from drones. The misalignment arises because the four instruments of the optical payload onboard the drone, namely, green (G), red (R), red edge (RE), and near-infrared (NIR), share the same acquisition device and thus work in time division. Consequently, they frame the same area from different positions along the trajectory. Although a preliminary coarse correction is capable of aligning the ground level, local changes in height originate local shifts among the bands. The proposed fine-alignment procedure is fully automated and does not require any additional information but is based on the unimodality of the optical sensors. The residue of the least-squares solution of the multivariate regression of G, R, and NIR toward RE yields the correction that is applied to G, R, and NIR to overlay them with RE. Experiments on a Tuscany vineyard are presented and discussed. The ripening cycle of vines is captured by the inflection point of reflectance spectra, whose calculation requires accurate alignment of the spectral components.