Registration Of Optical And SAR Image Based On Structure Features

Image matching technology has been widely used in the military field,major disaster monitoring and resource detection,and especially in the field of precision guidance,the technology has now occupied an important position.At present,various countries have conducted research on the guidance technology based on inertial guidance and image matching to improve the accuracy of target strikes to meters.In the process of image matching guidance,satellite SAR image and real-time SAR image are usually used to match.Due to the limitation of the existing technical conditions in our country,optical images are often used as the reference map and the SAR image as the real-time map.However,due to the difference in the imaging mechanism between the two,there is often a large difference in grayscale between the optical and SAR images.The traditional image matching algorithm based on grayscale information feature description is no longer applicable.Therefore,the research of optical and SAR image matching methods with high precision and real-time performance is of great significance to the development of precision guided weapons.To solve the above problems,in order to effectively avoid the influence of non-linear gray differences of the image on the matching,a feature description method based on structural information is used in this thesis to study the matching between optical and SAR images.Based on the research of SAR image line feature detection,the image target area calibration based on scene line features and the optical and SAR image matching methods based on the regional structure features are carried out.The specific research content of this article mainly includes the following three aspects:1.For the line feature detection of SAR image,this thesis first studies the traditional line feature detection methods,and discusses the effect of speckle noise in SAR images on line feature detection results.Based on this problem,this thesis combines the Ratio of Exponentially Weighted Averages(ROEWA)algorithm and LSD algorithm,proposes a line feature detection method based on ROEWA-LSD hybrid model,effectively overcomes the effects of speckle noise in SAR images and makes the simulation experiment.The experiment verifies that the algorithm can improve the accuracy of line feature detection and solve the problem of line feature fracture at the same time.2.To study the calibration of SAR image target areas.The structural information used in this thesis is mainly the geometric structure of the target area in the image.In order to achieve more accurate calibration of target feature areas in SAR images,based on the line feature detection method proposed in this thesis,a priori information is used to achieve accurate locate of the target area,and the regional growth algorithm is used to eliminate the Non-target feature area.3,for the optical and SAR image matching,using a rough to fine matching strategy,proposes an image matching method based on structural features.The rough matching is achieved by the Delaunay triangle mesh matching method,which effectively solves the influence of image rotation on the matching result.The template matching method based on Hausdorff measure is used to complete the fine matching.Based on the rough matching results,the fine matching search interval is further restricted,the search scope is narrowed,and the real-time performance of the algorithm is improved.In addition,in order to solve the problem of impact matching accuracy caused by nonlinear distortion in real-time graphs,this thesis studies a nonlinear image mapping method and verifies the effectiveness of the algorithm through simulation experiments.