Research On The Theory Of Mimicry Transformation Of Dual-model Infrared Image Fusion

The infrared intensity image mainly reflects the shape,brightness,and position information of the target;the infrared polarization image mainly reflects the edge and detail information of the target.The fusion of the two can more fully describe the scene information,improved the visual effects of the image,and has important applications in space detection,target recognition,security detection,high-performance air-day integrated network detection,and public safety video surveillance.How to better synthesize the complementary information from different source images to one image so as to improve the target detection and recognition level has important significance and value for practical application.With the continuous development of infrared detection technology,the scene information becomes more and more complex.The current infrared intensity and polarization image fusion algorithm cannot be dynamically adjusted according to the changes of the image difference characteristics,resulting in unsatisfactory or even ineffective fusion of partial image features.How to solve this problem has become the focus of today's infrared image fusion field.Based on the mimicry of bionics and the multi-mimicry process of mimicry octopus,this research proposes the theory of mimicry transformation of dual-modal infrared image fusion.In this research,it used the idea of mimic transformation to dynamically adjust the variable selection and combination methods in each variable group within the fusion algorithm,and provided the theoretical basis and the basis for the combination.Using image feature sensing in the framework of mimicry fusion,the perceived order of image features is input into the mimetic transformation framework.Finally,the optimal fusion algorithm was obtained.The proposed method given full play to the advantages of each variable group,significantly improves the fusion image quality,makes the main differences in theoriginal image features are well integrated,and improves the visual effect of the fused image.The main research contents of this article are as follows:(1)Analysis of multi-mimicry process and fusion process:By analyzing the reason of the mimicry process of mimicry octopus,the deep reason for its mimicry is obtained..By analyzing the general process of common image fusion algorithms,and explore the parts of the fusion process that can be independent of each other and have an impact on the fusion effect.By analyzeing the correlation between the mimicry process of the mimicry octopus and the image fusion process,and establish the correspondence between the two relationship.(2)In the framework of mimicry fusion,the variable group and its internal variables are selected and combined: In the process of image mimicry,the variable group was composed of the minimum transformation units.The different combinations of the variable groups act on the process of mimicry of the image fusion.The quasi-state fusion process was divided into four kinds of variable groups:multi-scale decomposition variables,fusion rule variables,fusion parameter variables,and fusion structure variables.By analyzing the fusion effect of the variables in the group for different image features,the fusion effect relationship between variables and image features was obtained,which provides theoretical support for the subsequent establishment of the theory of mimicey transformation.(3)Image Feature Perception in the Mimicry-Transformation Framework: Focusing on the relationship between the target scene and the image features,the probability distribution synthesis method was used to solve the problem of small sample size,perceptual feature probability and non-1.Through the feature extraction and quantification of the original image,the relationship between each type of image feature and the scene target was obtained,and its probability distribution was established.Using the probability theory,the probability distribution was transformed into a probability distribution,and the probability distribution was synthesized using a synthesis rule to obtain a synthetic distribution.This distribution was used to scene-sensing the scene information,to solve the fusion requirement problem and to subsequently establish a mimetic transformation principle and fusion.The dynamic adjustment of the algorithm lays the foundation.(4)The theory of mimicry transformation of dual-modal infrared image fusion: Firstly,the effect of different variables on the fusion of image features was analyzed,and the concept of comparative fusion degree was proposed,and then the effective fusion parameters and invalid fusion transformations for different image features are determined.The single mapping method was used to establish the mapping relationship between multi-variable groups and image features,and the multi-valued mapping relationship was further optimized.Mathematical modelling was used to model the obtained relationships using matrix methods to achieve the graphical representation.Into the functional representation,and then get the principle of mimetic transformation.The proposed theory could provided a basis for the choice of variables and the combination of them,avoiding the problem of the existing subjective choice of existing experience,and can improve the fusion effect.This not only solves the problem that the traditional fusion algorithm cannot be dynamically adjusted with the image feature transformation,but also solves the problem of lack of consideration of fusion requirements.