Research On Stabilizing Method Of UUV Visual Images When Approaching Operating Target

When approaching the operating target,the optical vision system of Unmanned Underwater Vehicle(UUV)is used to obtain the video,and then identify the size,type,position and attitude information of the target.This paper is motivated by the practical operation requirement that UUV autonomously identify the operating target,aiming at the problem of jitter in the video in the actual marine environment collected by UUV,the visual image stabilization method in the process of UUV approaching the operating target is carried out.Based on the camera imaging model,the transformation relationship between the four commonly used coordinate systems is deduced.The hardware configuration of the UUV optical vision system and the installation scheme of the camera on the UUV are discussed.Aiming at the jitter phenomenon of the video images collected by UUV,the characteristics of the underwater image as well as the reason and type of jitter video are discussed.By comparing the characteristics of the existing electronic image stabilization method in terms of accuracy and real-time performance,the specific scheme to stabilize the video images while UUV approaching the operating target is given.In aspect of motion parameter estimation,the improved motion transformation model of UUV optical visual images is proposed by comparing the characteristics and applicable scenarios of the existing image motion transformation model.The basic theory of two feature point extraction algorithms of Harris operator and SIFT operator are introduced in detail.By comparing the commonly used feature matching strategies,two different feature point matching strategies are given based on the characteristics of Harris operator and SIFT operator.Finally,based on the improved motion transformation model of UUV optical visual images,the motion parameter estimation method of the image sequence is deduced.In aspect of motion parameter compensation,Firstly,the basic principle and applicable range of the fixed reference frame compensation and the inter-frame compensation method are introduced.Secondly,aiming at the possible phenomenon of pixel overlap and the ‘empty' pixel phenomenon in which the pixel is not assigned that the classical forward mapping compensation method may causing,the inverse mapping compensation method is proposed.Finally,for the pixel point coordinates calculated by inverse mapping compensation method are floating point type,three subpixel interpolation methods,the nearest neighbor method,bilinear interpolation method and cubic interpolation method are compared in terms of accuracy and real-time performance,the bilinear interpolation method is finally chose to calculate the gray value of pixels in compensated frame image.In order to simulate the process of UUV autonomously identify the operating target in actual marine environment,the UUV pool simulation test platform is set up.The shaking image sequence captured by UUV approaching the operating target process is taken as the experimental sample,and the validity verification of the proposed algorithm is carried out.The result shows that the PSNR of the compensated image sequence is obviously improved than before,which means the proposed image stabilization algorithm is accurate and feasible.Applying the proposed image stabilization algorithm to jitter video acquired from actual ocean recovery when UUV is relatively static to the operating target can also achieve a good image stabilization effect.