Research On Remote Sensing Image Matching Based On Local Feature

Image matching is a very basic and critical issue in the field of digitalimage processing. Image matching is based on statistical decision theory, itbelongs to the area of computer vision and pattern recognition. The imagematching technology can be used to detect the concerned imageinformation in the original image. Especially in the field of remote sensing,image matching technology plays a very important role in target location,calibration and so on.Firstly, this paper introduces the development history and theprinciple of image matching technology and discusses the theoretical basisof image matching algorithm performance assessment. Secondly, in orderto enhance the performance of the traditional template matching algorithm,this paper creatively proposes a fast image matching algorithm based onlocal texture feature. In this paper we do a lot of experiments using theactual remote sensing image data to prove that the proposed algorithm hasimprovement in the performance compared to the traditional algorithm.Finally, we do some useful exploration about the combination of SIFT andRANSAC algorithm.Two main drawbacks of the traditional template matching algorithmare computationally intensive and low speed, with the development ofimage matching technology, many researchers had raised a lot of improvedalgorithm to reduce the computation of image matching algorithm toenhance the performance, however, these algorithms were at the expenseof the matching accuracy. This paper adopts the block partitioning strategy,texture feature analysis, and the FFT algorithm and integral image toimprove the performance of the conventional NCC algorithm.Experimental results show that the proposed algorithm is more robust andfaster than the conventional NCC algorithm.