Article Details

Abstract

Quality control of raw materials is crucial for ensuring the quality of tomato products. Traditional manual sorting is inefficient and its accuracy is easily affected by human factors, making it difficult to adapt to industrial large-scale production. Therefore, according to the actual needs of the tomato processing industry in Xinjiang, this study investigates the intelligent recognition of unqualified tomatoes based on deep learning and develops a supporting recognition system. This study constructs a dataset containing six types of tomato fruits through on-site photography and online collection, and expands it to 10,429 images through data augmentation. Four mainstream object detection models, namely Faster R-CNN, SSD, YOLOv8 and YOLOv13, are selected for comparative experiments, and YOLOv13 is finally determined as the optimal model. Experimental results show that the model achieves an mAP@0.5 of 94.8%, an inference speed of 32 FPS and a small object miss rate of 4.8%, and can accurately complete tomato localization, disease differentiation and qualification determination. In addition, a corresponding intelligent recognition system is developed to visualize the detection results, which supports four detection algorithms, namely YOLOv13, YOLOv8, Faster R-CNN and SSD, effectively replacing manual sorting and providing technical support and implementation solutions for intelligent quality control in the tomato processing industry.

Keywords

tomato processing; unqualified raw materials; intelligent recognition; deep learning; Faster R-CNN; SSD; YOLOv8; YOLOv13; object detection

References

1. [1] Zhao Jingyi Establishment of Tomato Quality Analysis and Non destructive Testing Model [D]. Shenyang: Shenyang Agricultural University, 2024
2. [2] Guo Yanying Development of Tomato Quality Rapid Detection and Visualization System Based on Hyperspectral Imaging Technology [D]. Yinchuan: Ningxia University, 2024
3. [3] Zhao Mingrui Research on Non destructive Testing of Tomato Fruit Quality Based on Hyperspectral Imaging Technology [D]. Shihezi: Shihezi University, 2023
4. [4] Li Chao, Kan Za, Zhang Ruoyu, etc Analysis and Research on Processing Tomato Sorting and Detection System [J]. Agricultural Machinery, 2009 (8): 86-87
5. [5] Liu Yande, Xu Hai, Sun Xudong, etc Non destructive testing of tomato maturity using near-infrared diffuse transmission spectroscopy [J]. Laser Technology, 2019, 43 (1): 25-29
6. [6] David I，Eisa B，Cedric O，et al. A Computer Vision System for Defect Discrimination and Grading in Tomatoes Using Machine Learning and Image Processing [J].Artificial Intelligence in Agriculture，2019，2（C）：28-37.
7. [7] Huang X Y，Pan S H，Sun Z Y，et al. Evaluating Quality of Tomato During Storage Using Fusion Information of Computer Vision and Electronic Nose [J/OL].Journal of Food Process Engineering，2018，41（6）：e12832.https://doi.org/10.1111/jfpe.12832.
8. [8] Guo Dechao, Rao Yuanli, Zhang Hao, etc Tomato comprehensive quality inspection method combining machine vision and spectral technology [J]. Food and Machinery, 2024, 40 (9): 123-130
9. [9] Sun Yuchao, Li Shouhao, Xia Xiubo, etc Using an improved YOLOv5s model to detect tomato fruit maturity and appearance quality [J]. Journal of Horticulture, 2024, 51 (2): 396-410
10. [10] He Tingting Tomato Appearance Quality Inspection Based on Machine Vision [D]. Jinzhong: Shanxi Agricultural University, 2023
11. [11] Lei M，Li S，Wu Y，et al.YOLOv13:Real-Time Object Detection with Hypergraph-Enhanced Adaptive Visual Perception [PP/OL].arXiv.https://arxiv.org/abs/2506.17733.

Funding

Innovation Training Program for Autonomous Region level College Students of Xinjiang Agricultural University in 2025 (S202510758002)

About the authors

Bao Wanting (2004-), female, Hui ethnicity, from Yinchuan, Ningxia, undergraduate student, research interests include artificial intelligence; Zhang Yunlong (2001-), male, Han ethnicity, from Urumqi, Xinjiang, undergraduate student, research interests include artificial intelligence; Wang Xintian (2004-), female, Han ethnicity, from Wusu, Xinjiang, undergraduate student, research interests include artificial intelligence; Zhang Yujing (2004-), female, Han ethnicity, from Urumqi, Xinjiang, undergraduate student, research interests include artificial intelligence; Corresponding author: Meng Xiaoyan (1978-), female, Han ethnicity, from Fukang, Xinjiang, doctoral degree, associate professor, master's supervisor, currently vice dean of the School of Computer and Information Engineering, research interests include knowledge graphs, computer application technology and IoT technologies.