| 引用本文: | 陆 棚,刘明堂,吴姗姗,等.基于HMM+LSTM算法的网纹蜜瓜数字孪生体生长模型设计[J].灌溉排水学报,2025,44(5):122-132. |
| LU Peng,LIU Mingtang,WU Shanshan,et al.基于HMM+LSTM算法的网纹蜜瓜数字孪生体生长模型设计[J].灌溉排水学报,2025,44(5):122-132. |
|
| 摘要: |
| 【目的】提高农业水资源利用效率，开展农作物生长过程全生命周期的数字孪生体构建，加快我国智慧农业进程、助力农民制订优化管理策略。【方法】以网纹蜜瓜为例，选取河南省花园口引黄灌区为典型研究区，在相应气候条件下开展网纹蜜瓜生长全过程室内试验，基于物联网技术的观测网络，获取了网纹蜜瓜生长过程各项环境指标和生长状态实时监测数据；采用3ds Max三维建模软件和Unity 3D可视化平台，开发了网纹蜜瓜数字孪生模型，采用隐马尔可夫（Hidden Markov Model，HMM）和长短期记忆网络（Long Short-Term Memory，LSTM）算法，构建了网纹蜜瓜生长过程智能化推演模型。【结果】模拟结果表明，网纹蜜瓜种、苗、花、叶、果不同生长周期的数字孪生体整体识别正确率较高，其中种周期与苗周期准确率为85.3%，网纹蜜瓜叶周期的准确率为78.6%，平均周期准确率为82.8%。【结论】本文提出的基于无线传感器网络的数据采集端系统、HMM+LSTM算法生成网纹蜜瓜孪生体三维生长模型，实现了智慧农业的精准、高效、非破坏性可视化全过程孪生模拟，可推广应用于其他农作物孪生体构建。 |
| 关键词: 数字孪生；网纹蜜瓜；隐马尔可夫模型HMM；长短期记忆网络算法LSTM；智慧农业 |
| DOI：10.13522/j.cnki.ggps.2024406 |
| 分类号: |
| 基金项目: |
|
| Design of netted muskmelon digital twin growth model based on HMM+LSTM algorithm |
|
LU Peng, LIU Mingtang, WU Shanshan, LI Bin, LI Shihao,
WANG Changchun, YANG Yangrui, JIANG Enhui
|
|
1. School of Information Engineering, North China University of Water Resources and Electric Power,
Zhengzhou 450045, China; 2. School of Electronic Engineering, North China University of Water Resources and Electric Power, Zhengzhou 450045, China; 3. Key Laboratory of Yellow River Sediment, Ministry of Water Resources,
Yellow River Institute of Water Resources Research, Zhengzhou 450003, China
|
| Abstract: |
| 【Objective】The purpose of this paper is to enhance agricultural water use efficiency, and to develop a digital twin system for simulating the entire growth life-cycle of crops, which holds significant importance for advancing smart agriculture in China and assisting farmers in formulating optimized management strategies. 【Method】Using netted muskmelon as a case, in the Yellow River Diversion Irrigation District of Huayuankou, Henan Province, we conducted controlled indoor experiments replicating local climatic conditions. An IoT-based monitoring network was employed to collect real-time data on environmental parameters and growth status throughout the cultivation process. The digital twin model was developed using 3ds Max for 3D modeling and Unity 3D for visualization, while the growth prediction model was built by integrating Hidden Markov Model (HMM) and Long Short-Term Memory (LSTM) algorithms. 【Result】Simulation results demonstrated high recognition accuracy across different growth stages: 85.3% for seed and seedling stages, 78.6% for leaf stage, with an overall average accuracy of 82.8%. 【Conclusion】The proposed system, combining wireless sensor networks with HMM+LSTM algorithms to generate 3D growth models of muskmelon digital twins, achieves precise, efficient, and non-destructive visualization of the entire growth process, and can be extended to construct digital twins for other crops. |
| Key words: digital twin; netted muskmelon; Hidden Markov Model (HMM); Long Short-Term Memory algorithm (LSTM); smart agriculture |
