Comparative analysis of image mosaicing techniques for aerial agriculture field imaging

UAVs that are outfitted with advanced image sensors offer a flexible and cost-efficient means of obtaining cohesive, inclusive, and precise visual depictions of regions. The obtained visual data, however, frequently comprises fragmented data as a result of the restricted perspective of individual images. To address this constraint, image mosaicing techniques are utilized to seamlessly merge various images, resulting in a thorough and uninterrupted depiction of the agricultural environment. The objective of this work is to evaluate and compare image mosaicing techniques, focusing primarily on agricultural datasets. It uncovers and exposes the constraints that exist in existing algorithms. The dataset of maize crop acquired from Phantom drones in the farms of Centralia, Missouri, USA, is utilized for this work. The Aerial images are used to evaluate distinctive feature-based and direct pixel-based mosaicing methods. The research analyzes four performance metrics, which consist of the Structural Similarity Index (SSIM) and Root Mean Square (RMS) error, together with Standard Deviation (SD) and CPU computational time. Analysis of Variance was conducted for the texture features of the generated mosaics. Several experimental findings show how the accuracy-performance-visual quality relationship affects the selection of suitable methods for real-time agricultural dataset. The research findings provide knowledge to select the best mosaicing solutions that will benefit crop health assessment and yield estimation applications.