Scholarlink

<p>Accurate estimation of fruit weight is critical for agricultural applications such as yield estimation, quality control, and market pricing. In this study, a&#xa0;nondestructive, image-based approach was developed to estimate the fresh weight of ‘Malaça’ (<i>Pyrus communis</i>) fruit, a&#xa0;local pear variety, via the Gaussian process regression (GPR) method. High-resolution images of the fruits were processed to extract dimensional features, which were then used as inputs for the GPR model. The model achieved a&#xa0;coefficient of determination (R<sup>2</sup>) = 0.7491, indicating that 74% of the variance in fruit weight could be predicted and that reasonable accuracy was achieved. Although this R<sup>2</sup> value is below the highest performance measure reported in the literature, the method is applicable for pear varieties with irregular geometric shapes. This study makes an original contribution by demonstrating the applicability of GPR-based weight estimation in local pear varieties. It is anticipated that future studies using larger sample groups, enriched feature sets, and different model architectures could improve estimation accuracy.</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

Automatic Weight Estimation of ‘Malaça’ Pears (Pyrus communis) via Gaussian Process Regression

  • 期刊论文

  • 发表日期：

    2026年2月12日

查看全文
  • İbrahim Hakkı Kadirhanoğullari

摘要

Accurate estimation of fruit weight is critical for agricultural applications such as yield estimation, quality control, and market pricing. In this study, a nondestructive, image-based approach was developed to estimate the fresh weight of ‘Malaça’ (Pyrus communis) fruit, a local pear variety, via the Gaussian process regression (GPR) method. High-resolution images of the fruits were processed to extract dimensional features, which were then used as inputs for the GPR model. The model achieved a coefficient of determination (R2) = 0.7491, indicating that 74% of the variance in fruit weight could be predicted and that reasonable accuracy was achieved. Although this R2 value is below the highest performance measure reported in the literature, the method is applicable for pear varieties with irregular geometric shapes. This study makes an original contribution by demonstrating the applicability of GPR-based weight estimation in local pear varieties. It is anticipated that future studies using larger sample groups, enriched feature sets, and different model architectures could improve estimation accuracy.