Study On Registration Technologies Of Different Species Remote Sensing Images

Remote sensing image registration is a vital step for remote sensing imageinformation fusion. Along with the development of remote sensing technologies, thedata amount of remote sensing images, the resolution of the sensors and the kinds of theimaging sensors have developed fast. In order to have a completely understanding of theobject, we must make full use of radiation properties of the different bands ofthe electromagnetic wave and information of different kinds of sensors. But first of allwe need ensure that the images should be co-registration geometrically. Firstly this paper introduces the concept and theory of the image registration, thegeneral image registration technologies and the geometrical transformation models. Andthen make an analysis and classification to the common technologies of the differentspecies remote sensing images and the imaging properties of the different kinds ofsensors.This paper introduces the imaging theory of remote sensing and the common imagepreprocessing technologies, and proposes a different species remote sensing imageregistration algorithm based on structure features. Structure features in this paper arecorner features extracted through FAST algorithm and the line support region (LSR)features. However, the imaging process of remote sensing images is very complicated,the variation of the radiation properties and geometrical properties are unsurely, thekinds of the sensors are diverse, and the application of the registration algorithm islimited. This paper proposes two different registration strategies based on differentimage properties and registration demand. Through experiments we get goodregistration results.In order to register the lidar data and the traditional remote sensing images, thepaper proposes a registration frame based on special ground objects. And then proposesa roof extraction algorithm in lidar altitude images based on infinite symmetricexponential filter. It proves that our algorithm have a good localization precisionthrough experiments.Then implements an image registration system based on the algorithm proposedand then overviews the work and the research results of this paper. At last we outlookthe research aspects in the future.