Research On Color Fusion Algorithm And Application Of Infrared And Low Light Level Images Based On Scene Recognition

Night vision sensors are important means to obtain effective information for night action,two common night vision images are infrared and low light level images.Color fusion can get natural feeling images and make target detection and recognition more accurate.Different from the existing color fusion methods based on fixed reference image,we proposed a fusion method that combines scene classification,fusion quality evaluation and color transfer,and researched on the application of color fusion method in bridge collision avoidance system.Based on the characteristics of infrared and low light level images,a scene classification method based on multi-spectral GIST features is proposed.Firstly,using infrared and LLL image to form the original color vector,then using the principal component analysis to reconstruct the vector and obtain a new four-dimensional color vector.Next calculating the local energy ratio of infrared and LLL images to divide the color vector into sparse sub grids and using the Gabor filters to get Multi-spectrum GIST features of the scene.Next using fuzzy support vector machine to finish scene recognition and using the color fusion quality evaluation method to find the best matched reference image in the image library of the corresponding category,finally we got the color fusion result of infrared and low light level image using color transfer method.Simulation experiments and effect evaluation on different data sets showed that the color fusion image is close to the natural color sensation,which made it easier to distinguish the target and improved the machine vision efficiency.Other researches are carried out on image database,image preprocessing,color fusion of sequence images with bridge collision avoidance system,which provided reference for the engineering application of color fusion technology.