3D Reconstruction Of Underwater Environment Based On Binocular Vision

In recent years,the development and utilization of marine resources has become increasingly fierce.Due to the difference between the underwater environment and the land environment,the underwater environment is more complex and unknown than land,there a series of technical problems in the exploration and development of the underwater environment.This paper is based on the binocular vision system,which is different from the previous application in the reconstruction of the terrestrial environment.It focuses on the study of underwater environment information,and the water information is collected through the image,and then preprocessed to further generate parallax with depth information to perform threedimensional reconstruction.In general,the underwater 3D reconstruction based on binocular vision is divided into the following steps: collecting and acquiring underwater images,calibrating the binocular camera,pre-processing the acquired image information,binocular image stereo matching to obtain parallax information,three-dimensional reconstruction of the subject.Firstly,the imaging model of the camera is explained,and the four coordinate systems common in binocular systems are explained in detail.At the same time,the transformation of the four coordinate systems under the mathematical concept is derived,then according to camera calibration method and illustrate.MATLAB is used to calibrate the binocular camera in the Toolbox tool based on Zhang’s calibration principle.The calibration environment is also in the same water tank used in subsequent experiments,ensure that the light refraction in the water is the same as the experiment.In the end the camera-related parameters required by the experiment are obtained.Secondly,it is difficult to deal with the difference between the underwater image and the land image and the low brightness and low contrast of the underwater image.A variety of image enhancement techniques are used for experimental comparison,and the processing method with the best effect is selected.After image enhancement,the defogging and restoration algorithms of underwater images are further compared and analyzed,and a defogging technique with edgepreserving filtering is introduced to make the image processing results reach a more ideal state.Thirdly,two common corner detection algorithms,Shi-Tomasi corner detection and subpixel corner detection,are experimented and the Surf feature extraction algorithm is reproduced as a theoretical basis for stereo matching.The experiment compares the local stereo matching algorithms NCC and Census based on feature points,and proposes a fusion stereo matching algorithm based on NCC and Census.Finally,the disparity map is obtained through experiments.Finally,the PSM stereo matching network is used for processing underwater images for the first time through experiments.The trained network is compared with the fusion stereo matching algorithm to select the best disparity map.3D reconstruction of disparity maps with depth information through Open GL,the reconstruction results highlight the contours of objects in the original image,but the overall 3D effect is poor,so the processing method based on PCL point cloud library is introduced to perform 3D disparity maps with depth information reconstruction,after a comparative experiment,the reconstructed image after point cloud information processing is better but there are discrete point clouds.This problem is solved after using least squares smoothing.The experimental results can intuitively reflect the threedimensional effect from multiple angles.And can better restore the underwater threedimensional environment.Therefore,this method is selected as the reconstruction algorithm required for the experiment.Three-dimensional reconstruction is achieved under the conditions of land and underwater contrast,insufficient underwater light,etc.The reconstruction results can better reflect the three-dimensional effect of the underwater environment.