Article Details

Abstract

To address the problems that hidden defects beneath the pomegranate peel and early-stage mildew are prone to missed detection and that visible-light detection has difficulty identifying internal damage, this study proposes a CSW-YOLO defect detection model based on infrared thermal imaging. Based on YOLOv10n, the model enhances the extraction and preservation of subtle defect features by replacing the C2f module in the backbone network with a C2f-GConv module, optimizing the SPPF module to form an SPPF-LSKA module, and replacing nn.Upsample with WaveletUnPool. Experimental results demonstrate that, while keeping computational overhead under control, the CSW-YOLO model achieves a precision of 0.881, a recall of 0.882, and an mAP@0.5 of 0.915, with overall performance superior to that of mainstream models such as YOLOv5, YOLOv8, YOLOv10, and YOLOv11. The model can effectively perform nondestructive automated detection of hidden defects beneath the pomegranate peel and early-stage mildew.

Keywords

CSW-YOLO; pomegranate defect detection; infrared thermal imaging

References

1. [1] Cai Zhonglei, Shi Ruiyao, Zhang Junyi, etc Early citrus rot detection method based on structured light reflection imaging and YOLOv7 CA [J]. Journal of Agricultural Machinery, 2025, 56 (8): 479-486495
2. [2] Wu Zhongxiao Research on Cherry Fruit Image Defect Detection Based on Digital Image Processing Technology [D]. Tianshui: Tianshui Normal University, 2025
3. [3] Li Xiao Research and Implementation of Appearance Defect Detection and Multi feature Grading System for Apple Sorting Line of S Company [D]. Jinan: Shandong University of Finance and Economics, 2025
4. [4] Zhang Yue Machine vision based method for identifying mature passion fruit and its quality grade [D]. Chongqing: Southwest University, 2025
5. [5] Wang D Y，Zhang M，Zhu Q B，et al. Beyond Human Vision: Highlighting and Detecting Inconspicuous Defects in Citrus Fruits Using UV-fluorescence Imaging and YOLO11 [J/OL].Postharvest Biology and Technology，2026，235：114190.https://www.sciencedirect.com/science/article/abs/pii/S0925521426000396?via%3Dihub.
6. [6] Cai Z L，Zhang Y Z，Li J B，et al. Synchronous Detection of Internal and External Defects of Citrus by Structured-illumination Reflectance Imaging Coupling with Improved YOLO v7 [J/OL].Postharvest Biology and Technology，2025，227：113576.10.1016/j.postharvbio.2025.113576.
7. [7] Molina M，Godoy J，Castro J W，et al. Apple Damages Classification: Using the Best Convolutional Neural Network to Discard low Surface Quality Fruit in Packing Plants [J/OL].Computers and Electronics in Agriculture，2026，244：111489.https://www.sciencedirect.com/science/article/abs/pii/S0168169926000840?via%3Dihub.
8. [8] Sheng X，Li Z Y，Wang D Y，et al. YOLO-ALDS: an Instance Segmentation Framework for Tomato Defect Segmentation and Grading Based on Active Learning and Improved YOLO11 [J/OL].Computers and Electronics in Agriculture，2025，238：110820.10.1016/j.compag.2025.110820.
9. [9] Li Lu, Sun Chaoqi, Zhou Yufan, etc Multi source image fusion recognition method for climbing shore grass crayfish based on FAS-YOLOv8n [J]. Journal of Agricultural Machinery, 2025, 56 (8): 526-534
10. [10] Tang Wenquan Research on nondestructive testing technology and sorting equipment of Century egg quality based on machine vision [D]. Wuhan: Huazhong Agricultural University, 2025
11. [11] Jie Dengfei, Shen Hengwei, Li Jiajun, etc Research on the Method of Pig Body Temperature Detection Based on Infrared and RGB Image Registration [J]. Chinese Journal of Animal Husbandry, 2025, 61 (5): 410-418
12. [12] Gao P，Zhong D，Qi Q，et al. FDM-YOLO: Real-time small-target UAV wildlife detection via attention-guided cross-modality fusion [J/OL].Ecological Informatics，2026，95，103697.https://doi.org/10.1016/j.ecoinf.2026.103697.
13. [13] Yan Y，Sheng Z Y，Gu Y，et al. Research note: A method for Recognizing and Evaluating Typical Behaviors of Laying hens in a Thermal Environment [J/OL].Poultry Science，2024，103（11）：104122.https://doi.org/10.1016/j.psj.2024.104122.
14. [14] Li J P，Zhang Y，Zhang Y，et al. MIF-YOLO: An Enhanced YOLO with Multi-Source Image Fusion for Autonomous Dead Chicken Detection [J/OL].Smart Agricultural Technology，2025，12：101104.https://doi.org/10.1016/j.atech.2025.101104.
15. [15] Sun R，Wang Q，Yu C，et al. A Multimodal Detection Method for Caged Diseased Hens Integrating Behavioral and Thermal Features via Instance Segmentation [J/OL].Computers and Electronics in Agriculture，2025，239(PartA)：110926.10.1016/j.compag.2025.110926.https://www.sciencedirect.com/science/article/abs/pii/S0168169925010324?via%3Dihub.
16. [16] Liu C，Hao B，Zheng Q，et al. An Improved YOLO-Based Algorithm for Infrared Small Object Detection [J/OL].Infrared Physics & Technology，2026，155：106443.https://doi.org/10.1016/j.infrared.2026.106443.
17. [17] Xue L L，Cong H Y，Xu D，et al. DBH-YOLO: A novel Algorithm for Surface Defects Detection of Postharvest blue Honeysuckle Berry Fruit [J/OL].Postharvest Biology and Technology，2026，233：114045.https://www.nstl.gov.cn/paper_detail.html?id = e7bf00ccd25868c6df72836d1b97bfe8.
18. [18] Yazdani M，Bao D，Zhou J，et al. Single-Wavelength Near-infrared Imaging and Machine Learning for Detecting Queensland Fruit fly Damage in Cherries [J/OL].Smart Agricultural Technology，2025，12：101090.https://doi.org/10.1016/j.atech.2025.101090.
19. [19] Canatan M，Muñoz-Carpena R，Boz Z. Continuous Surface Temperature Monitoring of Refrigerated Fresh Produce Through Visible and Thermal Infrared Sensor Fusion [J/OL].Postharvest Biology and Technology，2025，222：113354.https://www.sciencedirect.com/science/article/abs/pii/S0925521424005994?via%3Dihub.
20. [20] Li Junmao, Huang Zilu, He Dongli, etc Research on the Maturity Detection Algorithm of Greenhouse Tomatoes Based on YOLOv10n [J]. Journal of Inner Mongolia Agricultural University (Natural Science Edition), 2025, 46 (6): 68-75
21. [21] Zeng Kunfeng, Zhang Pengchao, Wang Lei, etc A lightweight tea pest and disease identification method based on improved YOLOv11n [J]. Journal of Agricultural Engineering, 2025, 41 (24): 220-229

Funding

2025 Guangxi University Young and Middle aged Teachers' Research Basic Ability Enhancement Project (Guijiao Research [2025] No. 1, 2025KY1409); School level Project of Guangxi Vocational and Technical College (Gui Vocational College [2023] No. 121, 231205)

About the authors

Liang Chenghe (1988-), male, Han ethnicity, from Nanning, Guangxi, is a senior information system project manager, lecturer, and master's student. His research interests include intelligent electromechanical equipment and computer vision; Corresponding author: Li Jiabao (1989-), male, Han ethnicity, from Nanning, Guangxi, lecturer, engineer, master's student, research direction: intelligent electromechanical equipment, mechanical design.