The Matching Of Multi-source High Resolution Image Based On Points And Lines

Matching technology based on multiple sources remote sensing data uses multi-scale information of the image to get more accurate, comprehensive and reliable description of target on the same place, which laid a good foundation for the further analysis in image processing. But at present the main algorithm based on single feature matching that can be easily affected by gray, scale, rotation angle and other factors of multiple source image data can’t meet the real-time feature extraction and the precision of matching requirements. In order to adapt to characteristics of multiple data, more scholars put forward the matching algorithm in combination of features that reduces the requirement of image feature by adding virtual feature. Even if there is a area with no obvious features, also get a better matching precision.This paper uses the matching method with a virtual point combination to line feature of multiple source image to make research and test that get the precision from coarse to fine based on feature extraction. The main content of this paper includes:(1) On the analysis of the traditional based on single feature matching algorithm and the existing problems, this paper introduces the present situation of matching technology characteristics and advantages, puts forward the technical route.(2) This paper discusses the characteristics of the classic point and line extraction algorithm, compares the advantage and disadvantage of each algorithm. Uses SIFT operator to extract point feature combination with the character of image data which has an advantage of multi-scale, angle rotation and strong robustness.In the line extraction, this paper introducing the traditional Hough transform principle and limitation puts forward the improved Hough transform based on binary of Canny edge image and sub-regional. Two groups experimental comparison analysis shows that this algorithm has a strong anti-interference ability, high positioning accuracy and improves line extraction speed.(3) This paper introduces the matching methods of virtual point combination with line feature which gets a precision from coarse to fine. This method makes the line extraction and matching based on sub-regional improved Hough transform on the basis of thick matching of SIFT operator, and introduces the concept of virtual point that gets high precision matching result in area being no obvious features.Using this matching method, this paper makes a matching test on two groups of Geo-eye and IKONOS, World-view and IKONOS, the result shows that the matching precision is sub pixel. Therefore, this algorithm has a certain application value for the image data with few features.