Analysis of Underwater Disparity Map and Parameter Optimization of SIFT Algorithm
Yongjie Pang, Di Wang · Advances in computer science research · 2015
The depth information of the underwater target is described with the disparity map.So the disparity map generation method is always a hotspot in binocular stereo vision research.The underwater target image stereo matching experiments were carried out by using the BM algorithm, SGBM algorithms and SIFT algorithm in the circulating water channel.Then the advantages and disadvantages of the disparity maps with the three kinds of stereo matching algorithms were analyzed.The analysis results of the disparity maps show that the SIFT stereo matching algorithm was more suitable for underwater stereo matching work.In order to obtain more feature points of underwater images, it is necessary to adjust SIFT algorithm parameter.Underwater image matching experiments were made to determine the appropriate Gaussian kernel parameter σ.The results illustrated that the improved SIFT algorithm has more feature points with σ=1.9.