Research On Multi-scale Scene Matching Based On NSCT

Scene matching is a kind of auxiliary navigation technology which relies on the advanced technology,such as sensor and image matching,to locate the aircraft accurately.The scene matching refers to an important image analysis and processing technique that identifies an image area from another corresponding scene area taken by another sensor or finds a correspondence between them.Scene matching as an image matching technology used in the navigation guidance of the last paragraph,used to improve the guidance accuracy and strengthen the guidance system autonomy.At present,scene matching technology has been widely used in remote sensing image processing,machine vision,missile guidance,medical diagnosis and other fields.With the rapid development of aviation and aerospace technology,the performance requirements of aircraft navigation become more and more high,especially in the positioning accuracy and real-time requirements,while the navigation system should also meet the high mobility,all-day work higher performance requirements.So far,real-time matching technology is the focus of research in recent years.This paper mainly studies the image matching algorithm based on multi-scale analysis.Based on the Non-subsampled Contourlet Transform(NSCT)as a tool for multi-resolution analysis,the image matching of heterogeneous images is studied for the imaging features of infrared and visible light images,and the corresponding algorithms are proposed.1.Based on the study of the traditional noise reduction algorithm,this paper presents an infrared image denoising and enhancement method combined with NSCT domain and airspace for the poor image quality of infrared image.Firstly,the image is transformed into the NSCT domain,and the image coefficients containing the noise are processed according to the 3 ? criterion to realize the NSCT domain infrared image noise reduction.And then in the airspace to establish a normalized non-complete beta model of gray beta,to achieve low-resolution and low-frequency image enhancement.The method can completely maintain the details of the image edge and achieve the effect of enhancing image quality.2.Based on the study of multi-scale image analysis theory,this paper proposes a multi-scale NSCT domain Krawtchouk invariant moment feature based on the multiscale analysis of the image with the visible image as the reference map and theinfrared image as the real-time graph.Scene matching algorithm which improves the matching algorithm from the three aspects of search space,search strategy,and feature space.The invariant moments are introduced into the feature space.The invariant moments are extracted by the invariant feature,and the feature search space is brought into the NSCT domain.Compress the search space,and use the improved genetic algorithm to improve the search speed.Experiments show that the proposed algorithm not only has higher matching precision and speed,but also has good robustness.3.Using the sparseness,multi-scale and anisotropy of the coefficients of NSCT,the SIFT feature has good stability in image description,and the improvement of SIFT feature for scene matching is proposed.A method combining NSCT with improved SIFT Matching algorithm which improves the robustness of the SIFT feature under the NSCT domain,and enhances the robustness of the algorithm from both the search space and the feature space.The experiment shows that the algorithm can resist some noise and geometric distortion.The content of this paper provides a new idea for solving the demand of high precision and high reliability in scene matching,which has certain practical value and can be applied to the view image as the visible image and the real-time image are the scene matching problem of infrared image.