Research On Recognition Of Infrared Target Based On UUV Under The Sea-sky Background

Unmanned Underwater Vehicle(UUV)is an indispensable and important tool for all the countries when they peacefully develop marine resources and defend the motherland in the seas.It not only plays an important role in the deep exploration and other civilian fields,but also plays an important role for military purposes as an maritime force multiplier.Infrared technology has been widely used in military reconnaissance,missile guidance and so on,because of its advantages such as being good at concealment,having long distance detection and working day and night.The research on recognition of infrared targets based on UUV under the sea-sky background combines the advantages of UUV and the advantages of infrared technology,so this research has important military significance and practical value in engineering.This paper is mainly aimmed to implement the processing of the images that the UUV shoots near the surface of the sea and the recognition of sea targets.Its main contents are as follows:(1)Firstly,the analyzing of the characteristics of the UUV near sea surface night infrared images studied in this paper has been finished,aimming at the characteristics of fuzziness and low contrast of the night infrared images an improved merge histogram stretching enhancement algorithm has been put forward and this method is based on the premise of “eat”deatails for image resolution.The experiments prove that this method can effectively improve the shortcoming of contour fuzziness and the contrast of the UUV near sea surface night infrared images and can make the image more clear.(2)Secondly,the other characteristic of the UUV near sea surface night infrared images is strong wave interference and this paper uses the idea of the sea-sky line extraction to realize the removal of wave interference.Through the experiments and researches of several traditional sea-sky line extraction methods and combining with the characteristics of the night infrared images,a sea-sky line extraction method based on line of entropy is put forward.The experimental results show that in this paper this method has better accuracy and applicability compared with other methods.In addition,combining with the characteristics of fuzziness and low contrast,an improved segmentation method based on the otsu method is proposed for the target area image which is acquired after the removal of wave interference of the night infrared images.(3)Thirdly,in order to discriminate the background of images is mountain or sea-sky,this paper respectively extracts the texture features of the original images and the images after the segmentation of the improved otsu method.Then according to the change of texture feature before and after segmentation determine the type of the image background.Finally,sea-sky background images is selected for the next feature extraction step.In this paper,the geometry features and HU invariant moments of the targets are extracted,after the analysis of the feasibility of the geometry features and HU invariant moments of the targets we can know that this group of features can realize representation and classification of targets.(4)Finally,this paper uses Support Vector Machine(SVM)to train the target features in the library of features and uses 10-fold cross-validation method to realize the selection of optimal parameters of the model.Then use the trained model to recognize the potential targets and the targets can be recognized and classified finally.It also has used the Matlab simulation software realized the building of the simulation platform of the recognition of UUV near sea surface infrared targets,and this platform makes the processing results of UUV near sea surface infrared images more intuitive,more vivid and more humanized.