The Registration Based On Features Of The Infrared And Visible Image

Infrared image reflects the heat radiation information of objects, and visible images mainly image according to the principle of light reflection, two images can describe objects from both heat distribution information and the light reflection information. The high precise registration can provide heat distribution of objects in the clear background, which has a very important practical significance and commercial value in locating the thermal fault of equipment and target in complex background.A key technology of equipment failures intelligent detection based on infrared and visible image fusion supported by National Natural Science Foundation of China has been researched in this paper. By researching the method to extract and describe the consistency features in infrared and visible image, this paper proposes a method to achieve a high precise matching between infrared and visible images.This work first studies the method of linear feature extraction and matching. The extraction method based on Hough Transform is adopted considering the features of infrared and visible image, then the describing and matching method is given next; afterwards this paper studies the commonly used feature points extraction method, and comprehensively compares performance of three kinds of feature points extraction algorithm from the aspect of rotation, scaling and brightness invariance. Then an algorithm of image registration based on Hough transform and SURF features is proposed. Firstly, linear features is used to accomplish the rough registration; then the image is divided into a few parts, and the SURF is used to do the precisely match sub-regionally; finally matching parameters are calculated and the registration is accomplished.The images downloaded from OTCBVS database and taken in laboratory are both used to test the proposed registration algorithm. Experiments based on MATLAB2010a software platform, show that the method can be used in the registration of infrared and visible images, comparing with traditional methods; this method can achieve higher accuracy in the presence of a large angle rotation and translation and get a better result when localized deformation or uneven illumination is occurring.