Abstract
Abstract. There are large geometrical deformations in SAR image, including foreshortening, layover, shade，which leads to SAR Image matching with low accuracy. Especially in complex terrain area, the control points are difficult to obtain, and the matching is difficult to achieve. Considering the impact of geometric distortions in SAR image pairs, a matching algorithm with a combination of speeded up robust features (SURF) and summed of normalize cross correlation (SNCC) was proposed, which can avoid the influence of SAR geometric deformation. Firstly, SURF algorithm was utilized to predict the search area. Then the matching point pairs was selected based on summed of normalized cross correlation. Finally, false match points were eliminated by the bidirectional consistency. SURF algorithm can control the range of matching points, and the matching points extracted from the deformation area are eliminated, and the matching points with stable and even distribution are obtained. The experimental results demonstrated that the proposed algorithm had high precision, and can effectively avoid the effect of geometric distortion on SAR image matching. Meet accuracy requirements of the block adjustment with sparse control points.
Highlights
The SAR sensor has side-view imaging method, resulting in the difference of SAR image deformation Large, there is a shadow, overlapping cover and other phenomena, and the lack of texture information in the mountain area, increasing the SAR image matching difficulty
In this paper,I adopt the combination of feature point extraction and summed normalized cross correlation (SNCC) to obtain evenly distributed matching points, and eliminates the error point pairs based on the two-way consistency constraint to improve the matching accuracy
The primary workhorse of speeded up robust features (SURF) algorithm is based on the integral image for speed up
Summary
The SAR sensor has side-view imaging method, resulting in the difference of SAR image deformation Large, there is a shadow, overlapping cover and other phenomena, and the lack of texture information in the mountain area, increasing the SAR image matching difficulty. In this paper,I adopt the combination of feature point extraction and summed normalized cross correlation (SNCC) to obtain evenly distributed matching points, and eliminates the error point pairs based on the two-way consistency constraint to improve the matching accuracy
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
