Research On 3D Reconstruction For Foreground Field Of View During AUV Recovery Docking

Autonomous Underwater Vehicle(AUV)as a powerful equipment for exploring marine environment and resources at present,can independently carry out a certain task with less human intervention,which greatly improved human's ability to explore the ocean.Due to the limited battery power carried by AUV itself,which affects its working time and distance,the research on AUV recovery docking has become a hot spot.This paper mainly studied the 3D reconstruction for the foreground field of view during the process of AUV recovery,proposed an underwater imaging model,completed the camera calibration,used the improved minimum spanning tree algorithm to achieve stereo matching,cooperated 3D reconstruction for the foreground field based on Point Cloud Library(PCL),and participated in the development of an AUV underwater recovery docking experiment system.Firstly,the paper analyzed the research status of AUV recovery and 3D reconstruction at home and abroad,expounded the overall scheme design and functional requirements of AUV recovery,and studied the design scheme of the 3D reconstruction for the foreground field of view.Secondly,the underwater imaging model was studied.Based on the principle of pinhole imaging,the binocular vision model in air and the influence of underwater environment on camera imaging were analyzed and an underwater imaging model considering refraction phenomena was proposed.The camera parameter calibration experiment was carried out,including taking target images on land and underwater.The calibration of internal and external parameters including refraction parameters was realized according to The plane calibration method of Zhang Zhengyou.Then an improved stereo matching algorithm was proposed to solve the problem that there were many weak texture and non-texture areas in the binocular image underwater.Based on MST,the algorithm combined CENSUS which robust to illumination noise,with MST to obtain the parallax map.According to the internal and external parameters,the parallax map was restored to a point cloud map containing 3D coordinates.The statistical filter was used to filter the noise.Based on the Open CV and the PCL,coordinates can be extracted in the point cloud image,which assisted in navigation and path planning to avoid collisions in the 3D reconstruction,Finally,the lake test was conducted.The transformation of the AUV and theproduction of the dock were introduced.The underwater software and the 3D reconstruction software were designed.The two key modules were experimental verification including binocular camera and the USBL,and the binocular camera was also tested in the pool to verify the feasibility of binocular visual positioning and the use of blue light for recovery.Lake test was conducted at Shazhou Lake,and the test showed that the development of AUV recovery docking system met the design.