Research Of Vegetation Image Mosaic Based On UAV

Image Stitching technology is computer vision, image processing and computergraphics research focus, it can be used to create a large visual, high-resolution images,are widely used in the field of virtual reality, the field of medical image processing,remote sensing technology and many other areas. Image mosaic is the same sceneoverlapping sequence of images stitched together to restore the image of the realscene large resolution. In recent years, with the rapid development of UAV technology,people have begun to use unmanned aerial vehicles equipped with cameras forlow-level image shooting, UAV technology of remote sensing images to make up forit more suitable for a small area of the image acquisition, the image clearer. Such asscenic shot by stitching panoramas more intuitive display to tourists; Meanwhile, ingeological disasters UAV technology applied very widely. Accordingly, for the UAVimage stitching related technology research has great practical value.This paper studies the image mosaic technology based on feature points consistprimarily of feature points selected image registration and image fusion of three parts,based on features of image stitching technology, followed by the problemsencountered in the panoramic image stitching process and use solution.The selected part of the feature points, first introduced the three corner-basedfeature point detection algorithm, SUSAN algorithm, Moravec algorithm and Harrisalgorithm. Harris corner detection algorithm, which describes the most detailed, it is astable and more mature algorithms. Then introduced more popular in recent years,SIFT (Scale Invariant Feature Transform) algorithm, the algorithm goes a step furtherin the Harris corner detection algorithm based on it is in the multi-scale extraction offeature points, feature point scale invariance, and because the algorithm for thecharacteristic point of the SIFT descriptor using a128-dimensional vector, therefore,the selected feature points having a strong unique. In this paper, UAV number ofimages, high resolution characteristics, taking into account the large number offeature points extracted by the SIFT algorithm, SIFT algorithm for UAV applications made small improvements, increased extreme points selected need compare thenumber of pixels to reduce the number of feature points, while not affecting thequality feature point extraction, thus reducing the number of comparisons in theregistration process of the feature point, and enhance the usefulness of the algorithm.In the image registration section, based on pixel square and bad characteristicsmatching and matching based on cross-correlation characteristics. Both feature pointmatching algorithm can be applied to the detected feature points matching process,such as the SUSAN corner the Moravec angle point and Harris corner. With punctual,thereby determining by comparing the pixel information of the feature points aroundthe correlation of the feature point, to achieve registration. For SIFT algorithm, whichis the Euclidean distance to determine the similarity between the two feature points,this is because the feature points extracted by the SIFT algorithm, including not onlythe position information of each feature point, and also uses a unique described in thefeature point matching process need only use the feature point position, scale,orientation, and the descriptor information on feature points can be paired, and theseinformation are generated in the feature point extraction process.In image fusion part, this paper analyzes the two commonly used image fusionalgorithm: the average method and the weighted average method.At the end of the article, Analysis of the Multiple image stitching process mayoccur in the accumulated error, the accumulated error includes the symmetry error andasymmetric error, they are a great influence on the quality of the final image mosaic,the accumulated error is dispersed to each stitching seams, effectively reduce theaccumulated error affect the quality of the panorama. In this paper, for the study offeature points matching algorithm, minimize the error in the process of graduallyexpand the stitching method to reduce the accumulated error on image quality.This paper studies the characteristics of the UAV images to study the existingimage registration algorithm based on feature points, and select the for UAV imagefeatures the stitching algorithm algorithm theory analysis and experimental simulationresults. In this paper, the final choice of the SIFT algorithm as a UAV imageregistration algorithm based on the unique characteristics of the UAV imageimprovements to make it more suitable for practical application. Then, panoramastitching process and the measures taken. Finally, this work is a summary of thesystem, and put forward their views on future research.