Research On Embedded Combination Of Multi-Algorithm Of Infrared Intensity And Polarization Image Fusion

Intensity and polarization detection technology of infrared are designed to image using different characteristics of the target respectively.The images obtained are highly complementary.The two techniques have been widely used in the field of military and civilian areas such as target detection,recognition and tracking.How to integrate information of two kinds of source images into a single image adequately so as to improve the level of target detection and recognition has great significance to practical application.The current single algorithm fusing two types of images focus on a single or several fusion features,which leads to fusion failure because it can not integrated important complementary information of source images fully.And duing to the lack of taking complementary of algorithms and combination sequences into account inadequately,the current combination algorithm has little enhance and even weakens some fusion features to fusion quality,thus falling short of the purpose of combinatorial optimization.This paper makes a deep research on the problem of multi-algorithm embedded combination fusion widely used,and explores how to get a better embedded combination.The main research is as follow:(1)Analysis on property of fusion algorithm: By the analysis of the relationship between the algorithm and image feature,algorithm theory and structure characteristics,this article selects different types of indicators for image features using infrared intensity and polarization images as the research object to measure and classify algorithms with methods of the Principal Component Analysis,Sigmoid Function Mapping and K-means clustering.Thecomplementarity between algorithms are analyzed,which provides the basis for the further analysis of algorithms and the selection of the complementary algorithm.(2)The relevant analysis on combination types between algorithms: The combination optimization problem of multi-algorithm is introduced based on combinatorial optimization principle.Different types of fusion algorithms are summarized,and the characteristics of each combination types are studied.The combination interface and combinatorial sequence are also analyzed subsequently.Combined with the specific fusion analysis,the advantages of embedded combination fusion and some related issues about embedded combination optimization are explored.All these studies provide the basis for the further analysis of embedded combination of algorithms.(3)The sequential decision problem of embedded combination fusion is introduced.Using the decision tree optimization principle as basis,this paper analyses the factors affecting the multi-algorithm embedded combination sequence and builds decision rules of combination sequence taking measurement of ‘non-distortion' condition as precondition,‘features separation' and ‘feature gain' as criterion.The effectiveness is proved with specific fusion algorithm analysis.(4)An embedded combination fusion method good for features separation and another one good for features enhancement using rule of ‘Saliency matching weighting' and another weak complementary combination algorithm based on multiple feature decision making are proposed for the three cases of embedded combination fusion.The effectiveness of the combination fusion method is verified with relevant experiments.