面向海参分布密度估计的水下观测机器人设计与实验

Design and Experimentation of Underwater Observation Robot for Estimating Sea Cucumber Distribution Density

摘要:

针对海参养殖领域传统人工盘点方法效率低、随机性强且覆盖率不足的问题，设计了一种可自主作业并能实时面积计算的水下观测机器人（ROV），旨在实现养殖区全覆盖扫描与精准生物量监测。首先，提出一种集视觉采集、运动控制、动态面积计算于一体的ROV系统框架；然后结合海参圈实际养殖场景设计基于PID的定高、定航的自主运动功能；接着依据运动学模型解算ROV位姿设计“几”字形路径规划算法，实现水下区域全覆盖，并依据相机标定参数动态计算相机扫描面积。实验结果显示，所设计的水下观测机器人垂直定位精度为±0.02m，面积计算误差小于2.31%，密度估算误差小于5%。该ROV以传感-运动-计算闭环设计，解决了水下机器人自主作业与实时面积反馈问题，为海参养殖提供了全自动、高可靠性的生物量监测方案。

Abstract:

Aiming to overcome the inefficiency, randomness, and insufficient coverage of traditional manual inventory methods in sea cucumber aquaculture, an autonomous underwater observation robot (ROV) with real-time area calculation capabilities was developed. The system was designed to enable comprehensive scanning of breeding zones and accurate biomass estimation. Firstly, an integrated ROV framework was proposed, combining visual acquisition, motion control, and dynamic area computation. Secondly, based on the operational requirements of sea cucumber ponds, autonomous motion functions, including depth-holding and course-keeping were implemented by using PID control algorithms. The ROV pose was determined through kinematic modeling, and a “crisscross” path planning strategy was employed to ensure complete seabed coverage. Real-time scanning area estimation was performed using calibrated camera parameters to enhance measurement accuracy. Experimental validation in simulated and real aquaculture environments demonstrated that the ROV achieved a vertical positioning accuracy of ±0.02m, an area calculation error below 2.31%, and a density estimation error under 5%. Furthermore, the system exhibited robust performance under varying turbidity conditions, maintaining stable vision-based detection even in low-visibility scenarios. The closed-loop control architecture, integrating perception, navigation, and computation, effectively addressed the challenges of autonomous underwater operation and real-time monitoring. Compared with conventional methods, the proposed system significantly improved inventory efficiency while reducing labor costs and human error. Future work would focus on multi-ROV collaborative operation and machine learning-enhanced sea cucumber recognition for large-scale aquaculture applications. This research contributed a fully automated, high-precision monitoring solution for sustainable aquaculture management.