Scholarlink

<p>The growing global appetite for fruits has led to a&#xa0;significant increase in their production in the agricultural sector. However, the process of packaging high-quality, undamaged fruits remains labor-intensive due to its reliance on manual methods. Conventional machine learning and deep learning approaches often struggle to detect and classify the varying sizes, shapes, and textures of fruits, resulting in suboptimal classification accuracy. To address these challenges, we introduce a&#xa0;novel neural network model for real-time fruit quality segregation (FQS-Net) using deep image feature extraction. Our approach leverages advanced multi-scale texture, shape, color, and semantic feature extraction, coupled with innovative fusion techniques, to effectively capture the intricate characteristics of fruits with high accuracy. The integration of <i>cross stage partial</i> (C3k2) and <i>cross stage partial with spatial attention </i>(C2PSA) blocks enhances the model’s focus on critical image regions, significantly improving detection accuracy and computational efficiency. These improvements help the network to manage various visual properties of fruits effectively, ensuring strong quality segregation. Experimental evaluations on the FruitsGB dataset demonstrate that our proposed model, FQS-Net, surpasses state-of-the-art models, achieving superior precision, recall, mAP50, mAP50-95, and inference speed. Furthermore, the model’s real-time applicability is validated through a&#xa0;GUI-based Web application, offering a&#xa0;transformative solution for automated fruit quality control. This work presents a&#xa0;robust and efficient tool for the food industry, promising enhanced productivity, packaging, and quality assurance.</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

Efficient Real-Time Fruit Quality Segregation Using Deep Image Features

  • 期刊论文

  • 发表日期：

    2026年2月17日

查看全文
  • Nitish Kumar,
  • Gaurav Mishra,
  • Saket Kumar,
  • Suraj Verma,
  • Rucha Gavas

摘要

The growing global appetite for fruits has led to a significant increase in their production in the agricultural sector. However, the process of packaging high-quality, undamaged fruits remains labor-intensive due to its reliance on manual methods. Conventional machine learning and deep learning approaches often struggle to detect and classify the varying sizes, shapes, and textures of fruits, resulting in suboptimal classification accuracy. To address these challenges, we introduce a novel neural network model for real-time fruit quality segregation (FQS-Net) using deep image feature extraction. Our approach leverages advanced multi-scale texture, shape, color, and semantic feature extraction, coupled with innovative fusion techniques, to effectively capture the intricate characteristics of fruits with high accuracy. The integration of cross stage partial (C3k2) and cross stage partial with spatial attention (C2PSA) blocks enhances the model’s focus on critical image regions, significantly improving detection accuracy and computational efficiency. These improvements help the network to manage various visual properties of fruits effectively, ensuring strong quality segregation. Experimental evaluations on the FruitsGB dataset demonstrate that our proposed model, FQS-Net, surpasses state-of-the-art models, achieving superior precision, recall, mAP50, mAP50-95, and inference speed. Furthermore, the model’s real-time applicability is validated through a GUI-based Web application, offering a transformative solution for automated fruit quality control. This work presents a robust and efficient tool for the food industry, promising enhanced productivity, packaging, and quality assurance.