Scholarlink

<p>This study explores the application of machine learning and genetic algorithm to model and optimize the spraying process of blueberry plants (<i>Vaccinium corymbosum</i> L.). The research aimed to determine the relationship between sprayer operating parameters (nozzle type, liquid pressure, and driving speed), weather conditions (temperature, humidity, and wind speed), and spray quality, specifically the coverage of various plant surfaces. Three ML methods, namely Gradient Boosting Regressor, Support Vector Regressor, and Extreme Gradient Boosting were evaluated to develop predictive models based on experimental data collected using water-sensitive papers. The results indicate that all models achieved high accuracy, with the SVR and XGBoost models being particularly effective, showing a MAPE error as low as 4% for certain surfaces. These models were then integrated into a genetic algorithm to identify optimal operating parameters that maximize total spray coverage under specific meteorological conditions. Optimization results revealed that XR nozzles generally achieved higher total coverage compared to AIXR nozzles. Furthermore, a driving speed of 5.5&#xa0;km/h was found to be optimal for both nozzle types across various conditions. This hybrid approach demonstrates significant potential for developing intelligent decision support systems in precision agriculture, enabling more efficient and sustainable use of plant protection products.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

The use of machine learning methods in modelling and optimization blueberry plants (Vaccinium corymbosum L.) spraying

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月23日

查看全文
  • Katarzyna Pentoś,
  • Beata Cieniawska,
  • Piotr Komarnicki

摘要

This study explores the application of machine learning and genetic algorithm to model and optimize the spraying process of blueberry plants (Vaccinium corymbosum L.). The research aimed to determine the relationship between sprayer operating parameters (nozzle type, liquid pressure, and driving speed), weather conditions (temperature, humidity, and wind speed), and spray quality, specifically the coverage of various plant surfaces. Three ML methods, namely Gradient Boosting Regressor, Support Vector Regressor, and Extreme Gradient Boosting were evaluated to develop predictive models based on experimental data collected using water-sensitive papers. The results indicate that all models achieved high accuracy, with the SVR and XGBoost models being particularly effective, showing a MAPE error as low as 4% for certain surfaces. These models were then integrated into a genetic algorithm to identify optimal operating parameters that maximize total spray coverage under specific meteorological conditions. Optimization results revealed that XR nozzles generally achieved higher total coverage compared to AIXR nozzles. Furthermore, a driving speed of 5.5 km/h was found to be optimal for both nozzle types across various conditions. This hybrid approach demonstrates significant potential for developing intelligent decision support systems in precision agriculture, enabling more efficient and sustainable use of plant protection products.