A multi-defect detection framework for sweet potato based on feature fusion and adaptive attention under complex postharvest conditions
文献类型：期刊论文
| 作者 | Guo, Xinyu3; Zhang, Jian3; Xiong, Haitao1; Zhang, Tao3; Yang, Ranbing3; Wang, Xufeng2 |
| 刊名 | JOURNAL OF STORED PRODUCTS RESEARCH
 |
| 出版日期 | 2026-03-01 |
| 卷号 | 116页码:14 |
| 关键词 | Sweet potato Defect detection Quality assurance Deep learning |
| ISSN号 | 0022-474X |
| DOI | 10.1016/j.jspr.2025.102937 |
| 通讯作者 | Yang, Ranbing(yangranbing@hainanu.edu.cn) |
| 英文摘要 | Sweet potato, as a major global economic and food crop, often suffers from surface defects due to its complex growth environment and harvesting processes. The long-term mixing of healthy and defective sweet potatoes during storage accelerates the physiological deterioration. This significantly reduces their edible quality and market value. To address this issue, this research presents SP-YOLOv11, a high-performance defect detection method designed to identify and grade four surface conditions of sweet potatoes in complex harvesting environments, including intact skin, minor defects, moderate defects, and severe defects. Specifically, first, a CPPA module is introduced, which employs a multi-branch feature extraction strategy to enhance the model's feature representation and detail-capturing ability. Next, an FDPN module is constructed, which, through multi-scale fusion and feature diffusion mechanisms, improves the adaptability to defects of different scales. Finally, an MSE-Detect module is introduced to focus on the key defect areas of sweet potatoes, further enhancing robustness in agricultural scenarios. To validate the effectiveness, eight ablation experiments are conducted. The results show that SP-YOLOv11 achieves the highest detection accuracy for sweet potato defects, with mAP@0.5 and mAP@0.5:0.95 reaching 98.50 % and 85.60 %, respectively. This research effectively overcomes the challenges of low efficiency, poor accuracy, and high damage rates in traditional defect detection methods. It significantly reduces the spread of diseases and physiological deterioration, while enhancing the post-harvest freshness retention and storage quality of sweet potatoes. The proposed approach provides technical support for their commercial processing and intelligent grading. |
| WOS研究方向 | Entomology |
| 语种 | 英语 |
| WOS记录号 | WOS:001659890800001 |
| 资助机构 | National Natural Science Foundation of China ; Key Research and Development Project of Hainan Province |
| 源URL | [http://ir.yic.ac.cn/handle/133337/42010]  |
| 专题 | 烟台海岸带研究所_中科院烟台海岸带研究所知识产出 |
| 通讯作者 | Yang, Ranbing |
| 作者单位 | 1.Chinese Acad Sci, Yantai Inst Coastal Zone Res, Yantai 264003, Peoples R China 2.Tarim Univ, Coll Informat Engn, Alaer 843399, Peoples R China 3.Hainan Univ, Haikou 570100, Peoples R China |
| 推荐引用方式 GB/T 7714 | Guo, Xinyu,Zhang, Jian,Xiong, Haitao,et al. A multi-defect detection framework for sweet potato based on feature fusion and adaptive attention under complex postharvest conditions[J]. JOURNAL OF STORED PRODUCTS RESEARCH,2026,116:14. |
| APA | Guo, Xinyu,Zhang, Jian,Xiong, Haitao,Zhang, Tao,Yang, Ranbing,&Wang, Xufeng.(2026).A multi-defect detection framework for sweet potato based on feature fusion and adaptive attention under complex postharvest conditions.JOURNAL OF STORED PRODUCTS RESEARCH,116,14. |
| MLA | Guo, Xinyu,et al."A multi-defect detection framework for sweet potato based on feature fusion and adaptive attention under complex postharvest conditions".JOURNAL OF STORED PRODUCTS RESEARCH 116(2026):14. |
入库方式： OAI收割
来源：烟台海岸带研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。