Abstract
The objective of this work is to create the 3D reconstruction of the archaeologicalobjects in underwater environment. The fusion technique is present, to obtainedthe 3D maps from the optical and acoustic systems. This work is divided intotwo main parts; First, we created the 3D reconstruction of the underwater scenesfrom the optic system by using the stereo cameras. Second, we created the 3Dinformaton of the underwater environment from the acoustic system by using themultibeam sonar. And we merge the two different types of map, i.e. from the opticand acoustic, which is all the more difficult for different task because of differentresolutions.The first part focus on the optical system used, to carry out the 3D reconstruc-tion by using the stereoscopic device. The underwater video and images sequencefor this work were acquired by divers in different underwater environment such asthe sea, the lake and the pool. First using a stereo camera to take a video of a cali-bration chessboard to calibrate the parameters of the camera, the intrinsic parame-ters are estimated for each camera, and then the external parameters are estimatedto determine the rotation matrix and translation vector of the stereo camera.The aims of this work is to create 3D reconstruction from multiple images.In each images pair,the features of interest are selected and matched across im-age pairs. An additional outlier removal step is performed based on the RANSACmethod. Triangulation of the inlier features from the 2D images space into a sparse3D points cloud is done by using a pinhole camera model and Euclidean distanceestimation. Then, the texture and rendering of the 3D stereo model are processed.The tempolral sequence of images is processed into a set of local 3D reconstructionwith estimate of coordinate transformation between temporally adjacent 3D localreconstruction by using the SFM method.The second part consists of fusing the 3D model obtained previously with theacoustic map. To align the two 3D models (optical model and acoustic model), we use a first approximate registration by selecting manually few points on each cloud.To increase the accuracy of this registration, we use analgorithm ICP (IterativeClosest Point).In this work we created a 3D underwatermultimodal map performedusingglobal 3D reconstruction model and an acousticglobal map.