Enhanced Method for Image Alignment of Urban Infrastructure Images by UAV Shooting

Relevance. Reducing the computational cost of image alignment procedures is an important field of research. The article considers the problem of processing urban infrastructure images obtained by UAV. The development of a high-speed method will make it possible to construct a digital map of an area by images from several UAVs in a limited time (ideally in real-time), which can be used to solve operational problems.Purpose of the work. Reducing the computational cost of the alignment procedure by preliminary estimating the parameters based on a limited sample of interest points. Methods used: analytical review of relevant scientific publications, experiment, algorithmization.The method for Image Alignment of Urban Infrastructure Images by UAV Shooting to make a digital map was developed to solve a scientific problem; the previously developed technique was enhanced; it reduced the computational complexity of image processing; the basis of the enhancement was the assumption that the transformation parameters range is limited; the advantage of the enhanced methodology were noted in the research.Results. An enhanced method of image alignment is proposed, which is characterized by shorter processing time than the original one. Experimental testing showed a reduction in time by half (from 50 s to 23 s), as well as a satisfactory result in combining 100 pairs of images.The work scientific novelty is determined by the author's approach to narrowing the range of the scaling coefficient and rotation angle for fragments matched of the alignment images. The approach is proposed for the first time.Theoretical significance. The assumption was confirmed that fragments of the image of urban infrastructure taken by UAV have a limited range of transformation of scaling and rotation.Practical significance. The research results can be used to make a machine vision system for digital map constructions in real-time. It provides a solution to operational problems of objects detecting and tracking the movement.

alignment, unmanned aerial vehicle, image, points of interest, geometric restriction. unique fragment, scale, rotate angle, computational cost