The Study Of Unmanned Aerial Vehicle Image Matching Method Base On SIFT Algorithm Improved By Block Model

UAV, with its flexible set route and time, low cost, phase of high, basically meet the requirements of precision, have been widely applied to monitoring, investigation, digital monitoring of urban, agricultural and forestry assessment, prediction and statistics. But accompanied by the large amount of UAV image data, manual processing complex, image matching as an important step in UAV image stitching, has become a hot spot of research, alongwith the domestic and foreign scholars for the matching algorithm research and improvement, so that the image matching technology has been developed, the formation of a lot of more mature matchingalgorithm.At present, SIFT algorithm is recognized as one of the most popular matching algorithm.This paper firstly introduces the current situation of image matching method, and then introduces the realization principle of SIFT algorithm. Finally, the SIFT algorithm is realized and improved. On the basis of the traditional SIFT implementation, using the image block strategy, based on the VC++6.0 platform programming SIFT algorithm and code encapsulation, and the use of multi process and multi thread to improve image matching speed. Comparative study based on Euclidean distance matching, the KD tree based matching, described the three key points matching method based on BBF, using RANSAC algorithm to remove the matching point of gross error, generation of corresponding image points and the rotation matrix. And the use of Anshun Guizhou airport UAV images for example, combined with the mainstream has been released by the SIFT operator and the time and space complexity to improve and evaluate the algorithm.