红树林智能种植机多传感器视觉系统设计与应用

Design and Application of Multi-Sensor Vision System for Mangrove Intelligent Planter

摘要: 为提升红树林种植作业的智能化水平，解决滩涂复杂环境下的作业难题，本方案设计并实现了一套多源感知与边缘决策系统。系统硬件集成多模态传感器，算法采用轻量化YOLOv10-MHSA模型进行实时目标检测，并基于LoRaWAN网络构建了远程监控与多机协同机制。田间试验验证了系统的有效性，实现了92.1%的坑洞检测准确率和96.1%的坑洞定位精度，结合改进遗传算法使种植空间利用率达到91.7%，有效种植点位数量提高25%。研究成果为红树林智能修复提供了可靠的技术方案。

Abstract: To enhance the intelligence of mangrove planting operations and address challenges in complex tidal flat environments, this solution designs and implements a multi-source perception and edge decision-making system. The hardware integrates multimodal sensors, employs a lightweight YOLOv10-MHSA model for real-time object detection, and utilizes a LoRaWAN network to establish remote monitoring and multi-device collaboration mechanisms. Field trials demonstrated the system’s effectiveness, achieving 92.1% accuracy in hole detection and 96.1% precision in hole localization. Combined with an improved genetic algorithm, it achieved a planting space utilization rate of 91.7% and increased the number of viable planting sites by 25%. These findings provide a robust technical framework for intelligent mangrove restoration.