The underwater targets were scanned and 3D reconstructed based on the autonomous underwater vehicle(AUV) platform equipped with a monocular camera and a line laser sensor. Aiming at the difficulty of extracting stripe coordinates in underwater laser images, an adaptive method was proposed for extracting laser stripe coordinates in the HSV space, and the 3D spatial coordinates of the target point were calculated by combining the laser triangulation model and the camera inverse projection matrix. Further, the AUV pose information was utilized to stitch multiple frames of point clouds, achieving the reconstruction of the complete 3D structure of the target object. By constructing the described algorithm framework described, AUV experiments were conducted to verify the feasibility of the method and the effectiveness of point cloud reconstruction under the pool environment.
为实现不同帧点云数据的拼接与对齐,需将相机坐标系下的点云转换至AUV坐标系下。随后,利用AUV的位姿信息对点云进行统一拼接与配准。通过测量相机坐标系原点在AUV坐标系中的位置,并结合旋转后的ZYX顺序欧拉角,可反推出从相机坐标系到AUV坐标系的旋转矩阵 R 和平移向量 t。利用该变换过程即可将点云数据统一转换至每个位姿的AUV坐标系下,其表达式如下:
其中,(XAUV,YAUV,ZAUV)表示空间点在AUV位姿坐标系的坐标。惯导输出的信息通常包括AUV的位置和姿态。对原始惯导数据进行处理,将其转换为包含时间戳和位姿信息的标准格式,包括:timestamp_usec、x、y、z、qx 、qy 、qz 、qw。timestamp_usec为微秒级unix时间戳,用于将点云数据与位姿信息进行匹配;x、y、z为AUV的坐标,四元数qx 、qy 、qz 、qw表示AUV的姿态。欧拉角、旋转矩阵、四元数都是用来描述旋转姿态的,它们之间可以相互转换。四元数作为一种紧凑且高效的旋转表达方式,不仅避免了欧拉角存在的奇异性问题(如万向节锁),也无需像旋转矩阵那样维护正交性约束,便于计算与插值,在三维空间变换中具有显著优势。对于AUV坐标系下的点 p =(XAUV,YAUV,ZAUV),当已知AUV在世界坐标系下的位姿(位置用 t 表示,姿态用四元数 q 表示),则其对应世界坐标系下的点 pw=(Xw,Yw,Zw)可表示为
CHENGuobang. Research on Multi-line Laser 3D Reconstruction of Underwater Scene Based on Binocular Vision[D]. Harbin: Harbin Engineering University, 2024.
WANGJiahuan. Research on Three-dimensional Reconstruction Method of Submarine Pipeline Based on Line Laser[D]. Harbin: Harbin Engineering University, 2022.
SUNQian, XUEQingsheng, ZHANGDongxue, et al. Research on the 3D Laser Reconstruction Method of Underwater Targets[J]. Infrared and Laser Engineering, 2022, 51(8): 234-240.
[19]
赵子毅. 基于线结构光的深海小区域三维重建[D]. 青岛: 青岛科技大学, 2020.
[20]
ZHAOZiyi. 3D Reconstruction of Deep Sea Area Based on Line Structured Light[D]. Qingdao: Qingdao University of Science & Technology, 2020.
[21]
XUEQingsheng, SUNQian, WANGFupeng, et al. Underwater High-precision 3D Reconstruction System Based on Rotating Scanning[J]. Sensors, 2021, 21(4):1402.
LIYixuan. A Semi Physical Simulation System Study on the Influence of Underwater Body Posture on 3D Reconstruction[D]. Harbin: Harbin Engineering University, 2024.
ZHANGTianchi, LIUYuxuan. Research Progress of Underwater Image Processing Based on Deep Learning[J]. Computer Science, 2024, 51(S1): 283-294.
[26]
孙志远. 基于结构光点云的水下目标三维重建技术研究[D]. 哈尔滨: 哈尔滨工程大学, 2024.
[27]
SUNZhiyuan. Research on Three-dimensional Reconstruction Technology of Underwater Targets Based on Structured Light Point Cloud[D]. Harbin: Harbin Engineering University, 2024.
YUShengchi, LIJiakang, XIONGXinquan, et al. Research on Fish Ranging Based on Line Laser Triangulation Ranging Method[J]. Fishery Modernization, 2024, 51(1): 80-89.