3D Reconstruction Of Underwater Environment Based On Biocular Stereo Vision

After rigorous research for many years, the theory of computer vision has been developing increasingly mature, and showing obvious or potential utility value in every research field. With the strengthening of Marine resources exploitation, the theory is also used to restore 3D model of underwater environment or target, which is mainly with the aid of binocular camera equipment that encapsulated in the underwater robot to acquire images. In this paper, on the basis of predecessors’ research, a stereo vision system has been set up to explore the 3D reconstruction scheme of the underwater environment at small scale and high density. The scheme mainly contains three aspects:(1) image acquisition; (2) feature point extraction and matching; (3) constructing three-dimensional contours.One kind of stereo vision camera, Bumblebee XB3, has been packaged to first engage in the underwater image collection at regular intervals as well as the pre-calibrated against distortion and misalignment. The corresponding epipolar lines of the stereo vision system were horizontal and on the same height.In this paper, image quality enhancement has been done before further operation combined with the homomorphic filtering and wavelet decomposition. Then a fast stereo matching algorithm has been implemented to accurately and efficiently construct the dense disparity map, i.e., the relative depth information of the stereo vision in the sea. Both the biocular consistency checking and uniqueness validation was taken to detect the occlusions, the unreliable matches, and to eliminate the false matches for the underwater vision system. Finally the 3D surface model can be completed on the basis of the depth map and the images after enhancement.It has been shown in the simulation experiments that the proposed model could get good performances in both robustness and effectiveness, and the colors and textures in the 3D reconstruction process could be corrected against the underwater light propagation. Due to the impacts on the stereo vision quality in the sea, a generic parameter-free enhancement method was presented to make a total abstraction of the image formation process at small scale, reduce the underwater perturbations, and correct the contrast disparities caused by the attenuation and backscattering.