Three Dimensional Reconstruction Algorithm Based On Binocular Stereo Vision In Underwater Environment

In the face of the gradual depletion of land resources today,people have an earnest keen on the ocean resources.The seabed contains rich energy,which have rich species and a huge number,for example,natural gas hydrate contained in the seabed has a great potential and is a rare new energy source,whose total number is nearly equivalent with twice the sum of the world's natural gas,oil and coal.Because of the undersea environment is more complex than that on land,in undersea exploration as well as mining process,the stereo vision technique is a necessary auxiliary means,and plays a key role in navigation,positioning,reconnaissance terrain and 3D reconstruction.Based on the binocular stereoscopic vision technology,this paper studies the three-dimensional reconstruction of the images collected in the underwater environment,which provides a strong support for the three-dimensional reconstruction of the underwater objects or scenes.The specific work of this paper is as follows:First of all,when taking photos of underwater scenes of the image,unlike the way of transmission in the air,as the propagation characteristics of light,the refraction phenomenon occurs when light propagates in different media,optical path changes,making the traditional imaging model do not apply to this article.Based on the imaging principle of the camera,the projection model and the back projection model of the underwater binocular stereoscopic vision are deduced in detail,which lays the foundation for the reconstruction of the 3D space points.Secondly,the SIFT feature points in the left and right views are extracted and the feature points in the left view are taken as the initial seed points.In this paper,the Mean Shift algorithm is introduced,and the left image is subjected to color-based images Segmentation,the pixels with the same color value are divided into the same region,and the candidate point diffusion criterion based on the Mean Shift algorithm is proposed as the constraint condition for the subsequent seed points diffusion.For each diffusion point,we determine the corresponding candidate region,and propose a Re-projection Similarity Distance method to determine three-dimensional points corresponding to thematching pairs,and finally we achieve the conversion of sparse matching to quasi-dense matching,and get more image feature information.Next,we get the 3D point cloud data from the quasi-dense matching.Due to the uneven distribution of the point cloud,this paper uses the Kriging interpolation algorithm to interpolate,and then uses the Poisson reconstruction algorithm to conduct reconstruction,generates a oriented patch structure from each 3D point,and finally we get the three-dimensional reconstruction result.Finally,this paper uses the binocular stereoscopic visual experiment platform of the laboratory to experiment,gives the relevant information of the experimental equipment,experiments on the underwater scene images using this algorithm,gives the experimental results,and analyses in detail for the experiment results,which provides a strong support for underwater reconstruction.