Researching On Underwater Stereo Matching Based On Depth Constraint

The ocean has immense potential resource and exploitation of marine resources is very important for relieving the stress of land resource shortage.The applications of underwater vision provides important technical support for water resource exploration and exploitation.We regard the application of underwater binocular stereo vision as the studying object and propose a dense stereo matching method that is used to underwater environment.We mainly do the following work:Firstly the linear imaging model of camera was introduced briefly,and it was extended to the underwater camera model based on this model.We analyzed parallel stereo system on land and underwater.After that,we introduced the basics of binocular stereo matching;We propose a underwater stereo matching searching region optimization method based on depth-constrained,binocular underwater model is the basic.The method can certain the constraint planes of underwater object through maximum imaging perspective of camera,further to depth range of underwater object.Finally,based on the underwater imaging model,we can identify range of horizontal Parallax and range of vertical distortion in the right correspond to any point in the left image,get the search area of depth constraints.The problem that epipolar constraint in the air does not meet to underwater stereo matching and matching algorithm in the air is not suit to water environment,proposing a semi global algorithm that the data items is improved.First we apply the depth constraints regional into AD and GRAD function that are often used in the air,then fusion both in some proportion;Within parallax search range,we used WTA strategy to get the best line deviation of vertical direction while getting the match cost at the same time,treating it as data items of energy function into half global algorithm for aggregate of match cost;Finally we use parabolic curve fitting to get dense underwater disparity map.