Development of an Integrated System for Forecasting Fruit Crop Yields Based on Multimodal Data and an Ensemble of Machine Learning Models

Abstract

The article presents a comprehensive system for forecasting orchard yields based on multimodal remote monitoring data. It combines convolutional neural networks for detecting flowers, ovaries, and fruits with an ensemble of linear and nonlinear models (multivariate regression, MLP, LSTM) for yield estimation. LASSO regression and SHAP analysis are used to interpret the results. The developed Python software enables full data processing, visualization, and saving of forecasts. The model achieves a determination coefficient of R2>0.85 and RMSE

Plant phenotyping relevance

果実園の収量予測を目的とするが、花・子房・果実をCNNで検出し、マルチモーダルデータを統合して収量を推定する取得・解析システムとPythonソフトウェアが中心であり、植物器官および収量形質の計測ワークフローに該当する。

Code and data availability

公開論文であることは確認できましたが、現在の公式API・許可済み取得経路では本文を自動取得できませんでした。

No evidence-backed public reproduction asset is currently recorded.