Methodology For Accurate Matching Of Feature Point In Remotely-sensed Imagery

The image matching is an important technology of image processing,it has been applied in many fields.In the field of remote sensing,remote sensing technology has been used in disaster,environment and resources,agriculture,forestry and so on,along with the rapid development of remote sensing technology,the application of remote sensing technology will expend in human life.In image processing and applications,we usually need the multi-source data and several images,such as using image fusion to improve the image’s resolution,using images overlapping to achieve the change detection,and the space stereo triangulation measurement.The image matching is a basic image processing technology,and the image feature point matching is a usual method to realize image matching.So,the image feature point matching method that is stability and good performance is important for image matching.The Scale-Invariant Feature Transform(SIFT)algorithm,because of its stability and good performance,it has been used in image matching.However,when the texture duplication is high,and the gray level difference is nonlinear,the feature point matching method based on feature description can’t meet the actual demand,and there are mistaking point matching.To achieve the correct feature point matching,this paper introduces new method to improve the accuracy of feature point matching.The details are as follows:(1)Considering the low matching point pairs and mismatching by taking the ratio of distance from the closest neighbor to the distance of second closest,this paper makes full use of the scale information and positioning information of feature points,using the scale ratio and coordinate offset threshold to achieve feature point matching.(2)Using the scale ratio and coordinate offset threshold to achieve feature point matching,and compare with SIFT in several image experiments,such as different time,scales and rotation changes,and fuzzy changes.The experimental results show that the introduced method can add the point pairs,and the root mean square error is lower.(3)To ensure accuracy of the feature point matching,the initial point pairs need to be processed.To eliminate the mismatching,this paper analyses the spatial relationship of the image feature point,and uses spatial relationship for image feature point matching.(4)The proposed method takes one-to-one corresponding point sets as an input,then uses spatial relationship to optimize the point sets.Comparing with GTM and RSOC method,experiments demonstrate the effectiveness of the introduced method.