A Tensor-based Algorithm For Underwater Optical Image Matching

In order to make better use of marine resources and strengthen marine defense,more and more new technologies have been applied to the vast sea.Different from the terrestrial environment,the water surface and the underwater environment and its special,a lot of application on the ground is not suitable for underwater technical requirements.In order to make better use of marine resources,the underwater optical image matching is studied in this paper.Based on the traditional algorithm,combined with the topological structure characteristics of the image and the iterative algorithm of tensor spectral radius,we can better achieve the matching of underwater images.For humans the forever visual image information received is the most direct and intuitive way,also in ocean exploration in the image is the most intuitive feeling can bring people,but the light propagation in water in different in air,so there will be a lot of optics limited areas and image in the water.For example,the existence of strong attenuation characteristics of light in water,and the influence of water bubbles and suspended solids in water on the propagation of light will cause many obstacles to optical imaging.In the AUV underwater reconnaissance,search aid,special operations,intelligence,military tasks and recovery,based on recognition,image matching and intelligent analysis is essential,and the traditional SIFT and SURF image matching algorithm is emphasized in matching point features and point features,because through the comparison of characteristics between points,can not reflect the characteristics of the structure of.Moreover,the feature points of underwater image are influenced by many factors.Especially under the artificial light source of submarine,it is easy to distort the information of feature points,which greatly affects the accuracy of matching.This paper studies the interest image for water extraction,using the corresponding feature tensor to establish two groups as well as the relationship between the visual image,the spectral radius of the iterative algorithm for solving symmetric tensor problems,so as to realize image matching based on tensor under water.Different from the existing algorithm through the structure feature where the spatial neighborhood relationship is established,which is intended in a variety of visual features of image combination,seeking an optimal solution to achieve the minimum difference in the combination of.The differential function is defined by the tensor constructed by the difference of the features between the matching graphs,and the matching is realized through the topology features of the local to the whole.Finally,the optimal matching solution is obtained by solving the tensor spectral radius by iterative algorithm,thus the matching of the image is realized.