Jiang Yun
College of Information Science and Engineering, Hohai University, Changzhou 213200, China.
Sensors (Basel). 2025 Mar 17;25(6):1850. doi: 10.3390/s25061850.
In this research, I introduce a water surface elevation measurement method that combines point cloud processing techniques and stereo vision cameras. While current vision-based water level measurement techniques focus on laboratory measurements or are based on auxiliary devices such as water rulers, I investigated the feasibility of measuring elevation based on images of the water surface. This research implements a monitoring system on-site, comprising a ZED 2i binocular camera (Stereolabs, San Francisco, CA, USA). First, the uncertainty of the camera is evaluated in a real measurement scenario. Then, the water surface images captured by the binocular camera are stereo matched to obtain parallax maps. Subsequently, the results of the binocular camera calibration are utilized to obtain the 3D point cloud coordinate values of the water surface image. Finally, the horizontal plane equation is solved by the RANSAC algorithm to finalize the height of the camera on the water surface. This approach is particularly significant as it offers a non-contact, shore-based solution that eliminates the need for physical water references, thereby enhancing the adaptability and efficiency of water level monitoring in challenging environments, such as remote or inaccessible areas. Within a measured elevation of 5 m, the water level measurement error is less than 2 cm.
在本研究中，我介绍了一种结合点云处理技术和立体视觉相机的水面高程测量方法。虽然当前基于视觉的水位测量技术主要集中在实验室测量或基于诸如水尺等辅助设备，但我研究了基于水面图像测量高程的可行性。本研究在现场实施了一个监测系统，该系统包括一台ZED 2i双目相机（美国加利福尼亚州旧金山的Stereolabs公司）。首先，在实际测量场景中评估相机的不确定性。然后，对双目相机拍摄的水面图像进行立体匹配以获得视差图。随后，利用双目相机校准的结果来获取水面图像的三维点云坐标值。最后，通过RANSAC算法求解水平面方程以确定相机在水面上的高度。这种方法特别重要，因为它提供了一种非接触式的、基于岸边的解决方案，无需物理水位参考，从而提高了在具有挑战性的环境（如偏远或难以到达的地区）中水位监测的适应性和效率。在5米的测量高程范围内，水位测量误差小于2厘米。
