摘要
为及时准确地获取波面信息,建立了基于双目立体视觉的波面重构流程,从二维波面图像中快速获取三维波面分布。使用双目相机拍摄的波面图像对作为原始数据完成相机参数标定,基于加速鲁棒性特征算法、金字塔搜索法和极线约束完成波面特征点的提取和立体匹配。最后通过立体矫正、视差图分析及图片后处理优化流程实现波浪场的三维点云重构,并选取重构区域作线性插值划分均匀网格,将三维点云投影至二维波面原始图像完成可视化。研究结果表明,在光照条件良好、风浪等级相对较大的情况下,双目立体视觉模型能够准确提取波面特征点,重建的三维点云能够再现波面,具有使用便捷且成本较低的特点,为后续做波浪等级分析及波高预报的相关研究奠定了基础。
To promptly and accurately obtain wave surface information,a wave reconstruction process based on binocular stereo vision is established to extract the three-dimensional wave surface distribution from two-dimensional wave surface images.The wave images captured by binocular cameras are used as raw data for camera parameter calibration.Speeded-up robust feature algorithms,pyramid search methods,and epipolar constraints are employed to extract and match wave surface feature points.Finally,the three-dimensional point cloud reconstruction of the wave field is achieved through stereo rectification,disparity map analysis,and post-processing optimization.A uniform grid is created by linear interpolation in the reconstruction area,and the three-dimensional point cloud is projected onto the original two-dimensional wave image for visualization.Research results indicate that under good lighting conditions and relatively high wind and wave levels,the binocular stereo vision model accurately extracts wave surface feature points,and the reconstructed three-dimensional point cloud faithfully represents the wave surface.This approach is convenient to use and has low cost,providing a foundation for subsequent studies on wave level analysis and wave height prediction.
作者
李蒙
刘曾
LI Meng;LIU Zeng(School of Naval Architecture&Ocean Engineering,Huazhong University of Science and Technology,Wuhan 430074,China;Hubei Provincial Engineering Research Center of Data Technology and Support Software for Ships(DTSSS),Wuhan 430074,China)
出处
《海洋工程》
CSCD
北大核心
2024年第5期157-164,共8页
The Ocean Engineering
基金
国家自然科学基金资助项目(12072126)。
关键词
双目立体视觉
波浪场重构
三维点云
加速鲁棒性特征算法
金字塔搜索法
binocular stereo vision
wave field reconstruction
three-dimensional point cloud
speeded-up robust feature algorithm
pyramid search method