Authors

Abstract

This study addresses the challenges of low-precision 3D visualization, unclear irrigation requirements, and inadequate smart decision-making in tea garden management. A twin data-driven 3D visualization and control system is proposed based on a six-dimensional digital twin framework, consisting of physical entities, virtual models, data connections, services, digital twin data, and decision mechanisms. First, real-time bidirectional data interaction is achieved through an OPC UA communication channel and multi-source sensor integration with a MySQL database. Second, parametric tea plant modeling in 3ds Max, combined with particle systems, dynamic shaders, ambient occlusion (AO), and level-of-detail (LOD) rendering, enables high-fidelity and dynamic 3D visualization. Finally, an AquaCrop-LSTM irrigation demand prediction model was developed by integrating the FAO Penman–Monteith method, the AquaCrop model, and a Long Short-Term Memory (LSTM) neural network. The complete system, deployed within Unity, forms a closed-loop architecture of perception, mapping, decision-making, and feedback. Experimental results show that the LOD strategy improves the frame rate by 121% while reducing vertex count by 93.7%. The AquaCrop-LSTM model achieves a mean absolute error (MAE) of 0.251 mm and an R² value of 0.927. Under a 30-user concurrent load test, the system maintained an error rate below 0.02%. These findings demonstrate that the proposed system provides reliable technical support for visual monitoring and efficient irrigation management in tea gardens.

Abstract in Chinese

针对茶园三维可视化精度低、喷灌需水量不明确及智能决策能力不足的问题，本文基于数字孪生六维架构（物理实体、虚拟模型、连接、服务、孪生数据与决策机制），提出一种孪生数据驱动的茶园三维可视化与系统设计方法。首先，基于OPC UA构建双向数据通道，结合MySQL集成多源传感器数据，为虚拟模型提供实时孪生数据支持。其次，基于孪生数据在3ds Max中进行茶树参数化建模，融合粒子系统、动态着色器、环境光遮蔽（AO）与多细节层次（LOD）技术，实现虚拟茶园的动态三维可视化。最后，集成FAO Penman-Monteith公式、AquaCrop模型与长短期记忆网络（LSTM），构建AquaCrop-LSTM需水量预测模型，并在Unity中封装形成“感知-映射-决策-反馈”闭环系统。实验表明：LOD使帧率提升121%，顶点数减少93.7%；AquaCrop-LSTM模型的平均绝对误差为0.251 mm，决定系数R²=0.927，精度显著优于对比模型；系统在30并发用户压力测试下错误率低于0.02%，响应稳定。该系统为茶园可视化监控与水资源高效管理提供了可靠技术支撑。