Tomato Growth Height Prediction Method by Phenotypic Feature Extraction Using Multi-modal Data  

基于多模态数据表型特征提取的番茄生长高度预测方法

作　　者：GONG Yu WANG Ling ZHAO Rongqiang YOU Haibo ZHOU Mo LIU Jie 宫宇;王玲;赵荣强;尤海波;周沫;刘劼(哈尔滨工业大学计算机科学与技术学院,黑龙江哈尔滨150006;智能农业技术与系统国家重点实验室,黑龙江哈尔滨150080;黑龙江省农业科学院园艺分院,黑龙江哈尔滨150040;哈尔滨工业大学人工智能研究院有限公司,黑龙江哈尔滨150000)

机构地区：[1]Department of Computer Science and Technology,Harbin Institute of Technology,Harbin 150006,China [2]National Key Laboratory of Smart Farming Technology and Systems,Harbin 150080,China [3]Horticultural Branch,Heilongjiang Academy of Agricultural Sciences,Harbin 150040,China [4]Harbin Institute of Technology Research Institute for Artificial Intelligence Inc.,Harbin 150000,China

出　　处：《智慧农业(中英文)》2025年第1期97-110,共14页Smart Agriculture

基　　金：中央高校基本科研业务费专项资金(2023FRFK06013);黑龙江省重点研发计划项目(2023ZX01A24);哈尔滨工业大学横向项目(MH20240081)。

摘　　要：[Objective]Accurate prediction of tomato growth height is crucial for optimizing production environments in smart farming.However,current prediction methods predominantly rely on empirical,mechanistic,or learning-based mod‐els that utilize either images data or environmental data.These methods fail to fully leverage multi-modal data to capture the diverse aspects of plant growth comprehensively.[Methods]To address this limitation,a two-stage phenotypic feature extraction(PFE)model based on deep learning algorithm of recurrent neural network(RNN)and long short-term memory(LSTM)was developed.The model integrated environment and plant information to provide a holistic understanding of the growth process,emploied phenotypic and temporal feature extractors to comprehensively capture both types of fea‐tures,enabled a deeper understanding of the interaction between tomato plants and their environment,ultimately leading to highly accurate predictions of growth height.[Results and Discussions]The experimental results showed the model's ef‐fectiveness:When predicting the next two days based on the past five days,the PFE-based RNN and LSTM models achieved mean absolute percentage error(MAPE)of 0.81%and 0.40%,respectively,which were significantly lower than the 8.00%MAPE of the large language model(LLM)and 6.72%MAPE of the Transformer-based model.In longer-term predictions,the 10-day prediction for 4 days ahead and the 30-day prediction for 12 days ahead,the PFE-RNN model con‐tinued to outperform the other two baseline models,with MAPE of 2.66%and 14.05%,respectively.[Conclusions]The proposed method,which leverages phenotypic-temporal collaboration,shows great potential for intelligent,data-driven management of tomato cultivation,making it a promising approach for enhancing the efficiency and precision of smart to‐mato planting management.[目的/意义]准确预测番茄的生长高度对优化智能农业中的生产环境至关重要。然而,目前的预测方法大多依赖于经验模型、机制模型或基于学习的模型,这些模型主要利用图像数据或环境数据,未能充分利用多模态数据,无法全面捕捉植物生长的各个方面。[方法]为了解决这一限制,本研究提出了一种基于深度学习算法的两阶段表型特征提取(Phenotypic Feature Extraction,PFE)模型,该模型结合了番茄植物的环境信息和植物本身的信息,提供了对生长过程的全面理解。PFE模型采用表型特征和时间特征提取器,综合捕捉两类特征,从而深入理解番茄植物与环境之间的相互作用,最终实现对生长高度的高精度预测。[结果和讨论]实验结果表明,该模型具有显著效果:在基于过去五天数据预测接下来的两天时,PFE-RNN(Phenotypic Feature Extraction with Recurrent Neural Network)模型和PFE-LSTM(Phenotypic Feature Extraction with Long Short-Term Memory)模型的平均绝对百分比误差(Mean Absolute Percentage Error,MAPE)分别为0.81%和0.40%,显著低于大语言模型(Large language model,LLM)模型的8.00%和基于Transformer的模型的6.72%。在较长期预测中,PFE-RNN模型在10天预测4天后和30天预测12天后的表现持续优于其他两个基准模型,MAPE分别为2.66%和14.05%。[结论]所提出的基于表型-时间协同的预测方法展示了其在智能化、数据驱动的番茄种植管理中的巨大潜力,是提升智能番茄种植管理效率和精准度的一种有前景的方法。

关 键 词：tomato growth prediction deep learning phenotypic feature extraction multi-modal data recurrent neural net‐work long short-term memory large language model 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] S641.2[自动化与计算机技术—控制科学与工程]