Research On Fusion Model Of Infrared And Visible Video Mimicry Driven By Difference Features

The fusion of infrared and visible video is the process of complementary fusion of infrared video radiation information and visible video details,which can effectively improve the quality of video fusion and precision detection performance,it has great application value in the fields of medical image,security monitoring,flood control and disaster reduction,national defense and so on.However,in the traditional video fusion technology,the fusion algorithm and the fixed model can not change dynamically with the change of the difference feature,which causes the problem of poor fusion effect,the intelligent detection effect of the system is seriously restricted.Therefore,in this paper,the research of intra-frame difference feature and the mapping relationship between intra-frame difference feature and class set of fusion algorithm are discussed,finally,a difference-driven pseudo-fusion model of infrared and visible video is constructed,which can break through the bottleneck of poor fusion effect caused by the fixed fusion algorithm,rule,parameter,structure and so on.The main content of this paper is as follows:(1)The research of intra-frame and inter-frame differences in video.Firstly,by analyzing the mechanism of infrared and visible video imaging,the different feature types in the fusion process are defined,and then the connotation of inter-frame difference and intra-frame difference features are described quantitatively by using the formula At last,the variation characteristics of intra-frame difference feature in dynamic scene are illustrated by using the variation characteristic curve of difference feature amplitude,which lays a foundation for effective driving of video mimicry fusion model.(2)The research on the mapping relationship between intra-frame difference features and the class set of fusion algorithms.Firstly,according to the fusion effect of typical fusion algorithm,the difference feature and fusion algorithm class set for infrared and visible video fusion are constructed Then,by constructing the fusion validity and distribution synthesis rules,the mapping relationship between the difference features and the class set of fusion algorithms is established,the validity and feasibility of the fusion validity of cosine similarity are verified.(3)The research on video mimicry fusion model of infrared and visible light.Firstly,the whole framework of the video mimicry fusion model driven by difference features is built,and secondly,an adaptive fusion validity weight calculation method based on entropy weight method is proposed to rank the fusion algorithm,then,from the subjective and objective point of view,the fusion results are evaluated and the objective evaluation indexes are synthetically constructed Finally,the feasibility of the proposed method is verified by experiments.