摘要
[目的]实现番茄内外部品质同时快速准确测定,提高番茄的分级效率和品质。[方法]基于机器视觉和光谱技术,提出一种结合外部品质和内部品质的番茄综合品质分级方法。通过对YOLOv8模型进行4个方面的优化(轻量化卷积、小目标检测层、CBAM注意机制和损失函数)完成外部缺陷检测,结合果形指数和番茄尺寸完成外部品质分级。通过预处理方法、特征提取方法和改进粒子群优化的最小二乘支持向量机完成番茄内部品质分级。通过试验对所提分级检测方法的性能进行分析。[结果]所提方法可以实现番茄综合品质检测,具有较高的准确性和效率。外部品质分级准确率>93.00%,内部品质分级准确率>86.00%,融合品质分级准确率>96.00%,平均分级时间<0.25 s。[结论]结合机器视觉和光谱检测技术可以实现番茄综合品质的快速、无损和准确评估。
[Objective]To realize rapid and accurate measurement of both internal and external quality of tomatoes,and improve the efficiency and quality of tomato grading.[Methods]Based on machine vision and spectroscopy technology,proposed a tomato comprehensive quality grading method which combined external and internal quality.By optimizing the YOLOv8 model in four aspects(lightweight convolution,small object detection layer,CBAM attention mechanism,and loss function),external defect detection was completed,and external quality grading was achieved by combining fruit shape index and tomato size.Complete tomato internal quality grading through preprocessing methods,feature extraction methods,and improved particle swarm optimization using least squares support vector machine.Analyzed the performance of the proposed grading detection method through experiments.[Results]The proposed method could achieve comprehensive quality testing of tomatoes with high accuracy and efficiency.The accuracy of external quality grading>93.00%,the accuracy of internal quality grading>86.00%,the accuracy of fusion quality grading>96.00%,and the average grading time<0.25 s.[Conclusion]Combining machine vision and spectral detection technology can achieve rapid,non-destructive,and accurate evaluation of tomato comprehensive quality.
作者
郭德超
饶远立
张豪
李春峰
赵强
GUO Dechao;RAO Yuanli;ZHANG Hao;LI Chunfeng;ZHAO Qiang(Guangzhou University of Chinese Medicine,Guangzhou,Guangdong 510006,China;Guangzhou Center for Disease Control and Prevention,Guangzhou,Guangdong 510440,China;Guangdong University of Technology,Guangzhou,Guangdong 510006,China;South China Agricultural University,Guangzhou,Guangdong 510642,China)
出处
《食品与机械》
CSCD
北大核心
2024年第9期123-130,共8页
Food and Machinery
基金
广东省教育厅科研项目计划课题(编号:21GZJY675032)
广州市哲学社科规划课题(编号:2023GZGJ64)。
