摘要
针对传统的大尺寸船体曲板接触式样箱测量方法难以实现自动化、测量精度低,且不能量化测量误差的问题,提出了一种非接触式基于计算机视觉的船体曲板摄影测量方法.利用摄像机标定算法实现对摄像机内参数和外参数的标定.为得到船体曲板表面的更多特征信息,提出了在船体曲板表面投射结构光的摄影测量方法;并基于AKAZE特征提取算法,将KNN机器学习算法与RANSAC算法相结合进行特征匹配与优化,利用基于特征的准稠密扩散算法实现船体曲板的高精度三维重建.搭建基于计算机视觉的双目摄影测量系统,进行摄像机标定实验和船体曲板测量实验,并进行精度分析.试验结果表明,相比于传统的船体曲板测量方法,在测量场景不小于3000 mm×3000 mm,船体曲板尺寸为1500 mm×2000 mm范围内,测量误差小于1 mm,满足船舶制造的精度要求,实现了大尺寸船体曲板相对高精度测量.
Aiming at the problems that the traditional measurement method of large-size hull curved plate contact pattern box is difficult to realize automation,the measurement accuracy is low,and the measurement error cannot be quantified,a non-contact photogrammetry method of hull curved plate based on computer vision was proposed.The camera calibration algorithm was used to calibrate the internal and external parameters of the camera.In order to obtain more characteristic information of the curved hull plate surface,a photogrammetry method of projecting structured light on the curved hull plate surface was proposed.Based on AKAZE feature extraction algorithm,KNN machine learning algorithm and RANSAC algorithm were combined for feature matching and optimization,and feature-based quasi-dense diffusion algorithm was used to realize high-precision 3D reconstruction of curved hull plate.Based on computer vision,a binocular photogrammetry system was set up,camera calibration experiment and hull curved plate measurement experiment were carried out,and the accuracy was analyzed.The test results show that,compared with the traditional measurement method of curved hull plate,the measurement error is less than 1mm when the measurement scene is not less than 3000 mm×3000 mm and the size of curved hull plate is 1500 mm×2000 mm,which meets the precision requirements of shipbuilding and realizes the relatively high precision measurement of large-sized curved hull plate.
作者
胡得林
胡勇
于洋
何朋朋
俞峰
HU Delin;HU Yong;YU Yang;HE Pengpeng;YU Feng(School of Naval Architecture,Ocean and Energy Power Engineering,Wuhan University of Technology,Wuhan 430063,China;Jiangnan Shipyard(Group)Co.Ltd.,Shanghai 201913,China)
出处
《武汉理工大学学报(交通科学与工程版)》
2022年第4期638-643,共6页
Journal of Wuhan University of Technology(Transportation Science & Engineering)
基金
国家自然科学基金(51779200)
中央高校基本科研业务费(215202001)。
关键词
大尺寸船体曲板
计算机视觉
摄影测量
摄像机标定
特征匹配
三维重建
large-size curved hull plate
computer vision
photogrammetry
camera calibration
feature matching
3D reconstruction