Scholarlink

<p>Accurate postharvest potato grading is vital for maintaining market value and ensuring uniform quality. Conventional grading methods, reliant on human inspection, are often inconsistent, labour-intensive, and prone to error. In this study, a real-time automated potato grading system was developed using machine vision integrated with support vector machine (SVM) classifier for size-based classification. The system comprises a conveyor unit, a controlled lighting chamber, a web camera, a processing module, and a servo-driven ejection mechanism. A total of 180 potatoes (60 each of small, medium, and large) were analyzed. From captured images, nine morphological features were extracted, among which surface area, major axis, minor axis, equivalent diameter, and perimeter showed statistically significant differences across categories, as determined by analysis of variance (ANOVA). Also, the Principal Component Analysis (PCA) confirmed effective class separation. The SVM classifier achieved an accuracy of 96% for the testing dataset. When deployed in real-time ejection operation, the system achieved an ejection accuracy of 98.1% across mixed-size classes. However, when assessed under real-time operation from feeding to ejection, actual grading efficiencies ranged from 71 to 86% for mixed-size potatoes. The suggested method, which utilizes machine vision and the support vector machine algorithm for potato grading, demonstrates excellent reliability and efficiency.</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

Real-Time Automated Potato Grading Using Machine Vision and Support Vector Machine Algorithm

  • 期刊论文

  • 发表日期：

    2026年4月28日

查看全文
  • Abdullah Al Masum,
  • Md. Mahedi Hasan Himel,
  • M. Mirazus Salehin,
  • Kazi Shakibur Rahman,
  • Al Amin,
  • Sahabuddin Ahamed,
  • Md. Rostom Ali,
  • Anisur Rahman

摘要

Accurate postharvest potato grading is vital for maintaining market value and ensuring uniform quality. Conventional grading methods, reliant on human inspection, are often inconsistent, labour-intensive, and prone to error. In this study, a real-time automated potato grading system was developed using machine vision integrated with support vector machine (SVM) classifier for size-based classification. The system comprises a conveyor unit, a controlled lighting chamber, a web camera, a processing module, and a servo-driven ejection mechanism. A total of 180 potatoes (60 each of small, medium, and large) were analyzed. From captured images, nine morphological features were extracted, among which surface area, major axis, minor axis, equivalent diameter, and perimeter showed statistically significant differences across categories, as determined by analysis of variance (ANOVA). Also, the Principal Component Analysis (PCA) confirmed effective class separation. The SVM classifier achieved an accuracy of 96% for the testing dataset. When deployed in real-time ejection operation, the system achieved an ejection accuracy of 98.1% across mixed-size classes. However, when assessed under real-time operation from feeding to ejection, actual grading efficiencies ranged from 71 to 86% for mixed-size potatoes. The suggested method, which utilizes machine vision and the support vector machine algorithm for potato grading, demonstrates excellent reliability and efficiency.