Application Of Remote Sensing Image Registration Based On Dual Channel Convolution Neural Network In Ship Navigation

With the rapid development of shipbuilding industry and marine economy,the importance of maritime traffic safety can not be ignored.How to avoid collision,plan route and safely berth according to the surrounding environment has become a very important preventive measure.In ship image system,image registration technology is an effective way to generate highresolution panorama,and it is also an important measure to assist crew to drive safely and strengthen ship information construction.Remote sensing image refers to the image with ground target characteristics obtained by aircraft or satellite imaging system.Remote sensing image registration is the process of finding the best alignment between image pairs captured at different times or by different sensors.At present,remote sensing image registration technology has been widely used in ground target recognition,urban and geographical change evaluation and other military and civil fields.In order to meet the needs of remote sensing image registration in various fields,this paper proposes a remote sensing image registration method based on dense structure improved dual channel neural network,and another remote sensing image registration method based on full convolution neural network and k-nearest neighbor ratio algorithm.1.This paper proposes a remote sensing image registration method based on dense structure improved dual channel neural network.Firstly,the dense structure improved dual channel convolution neural network model is used for feature extraction of the input image,and then the random consistent point drift algorithm improved by particle swarm optimization is used for feature matching to obtain the affine transformation coefficient.Finally,the image to be registered can be transformed according to the coefficient to achieve the purpose of registration.2.This paper proposes a remote sensing image registration method based on full convolution neural network and k-nearest neighbor ratio algorithm.In this paper,a coarse to fine registration method is proposed.In the coarse registration stage,the full convolution neural network is used to extract the features of the input image,and then the nearest neighbor distance ratio algorithm is used to coarse match the features,and finally an approximate transformation matrix is obtained.In the stage of fine registration,firstly,the image features are extracted by the improved convolutional neural network based on jump connection,then the affine transform coefficients are obtained by the combination of approximate transform matrix and k-nearest neighbor ratio algorithm.Finally,the image to be registered can be transformed according to the coefficients to achieve the purpose of registration.Experiments show that the registration accuracy of this method is higher than that of some traditional algorithms,and the proposed method can effectively improve the registration accuracy of remote sensing images with significant geomorphic differences or multi view remote sensing images.