摘要
目前农产品无损检测方法单一,导致分拣分级精度不高,亟须提升批量动态检测精度。针对梅州蜜柚检测,采用光谱检测、机器视觉、动态称重和人机交互等技术,设计了一款智能驱动无损动态检测装置。阐述了装置关键结构参数与选型、搭建方法与工作原理,并开展了性能测试。测试分析结果表明,该装置可采集和融合柚果特征波长、图像特征、和动态质量等多特征信息,并自主完成多源异质信息融合,通过柚果内部品质深层解析模型——广义神经网络模型,再经样本训练、规则判别、算法寻优和深度学习,快速完成内部品质指标的精确预测和验证。说明所设计的智能驱动无损动态装置可以快速、高效和准确地完成柚果品质检测,满足性能需求。
At present,there is not many non-destructive testing methods for agricultural products,resulting in low sorting and grading accuracy.It is urgent to improve the accuracy of batch dynamic testing.By adopting spectral detection,machine vision,dynamic weighing,and human-machine interaction,a smart driven non-destructive dynamic testing device was designed for Meizhou honey pomelo testing.The key structural parameters,type selection,construction methods and working principles of the device were elaborated,and performance testing was conducted.The test analysis results show that the device can collect and integrate multiple feature information such as pomelo fruit feature wavelength,image feature,and dynamic quality,and independently complete multi-source heterogeneous information integration.Through the deep analysis model-generalized neural network model,and through sample training,rule discrimination,algorithm optimization,and deep learning,the accurate prediction and verification of internal quality indicators can be quickly completed.The intelligent driving non destructive dynamic device designed can quickly,efficiently,and accurately complete the quality inspection of pomelo fruit,meeting performance requirements.
作者
孙潇鹏
吴良军
车志
马创立
郭海龙
Sun Xiaopeng;Wu Liangjun;Che Zhi;Ma Chuangli;Guo Hailong(Department of Automobile and Engineering Machinery,Guangdong Communication Polytechnic,Guangzhou 510650,China;Guangna New Energy Technology(Guangzhou)Co.,Ltd.,Guangzhou 510530,China)
出处
《机电工程技术》
2023年第4期16-20,共5页
Mechanical & Electrical Engineering Technology
基金
广东省教育厅特色创新项目(自然科学类)(编号:2018GKTSCX080)
广东省普通高校特色创新类项目(编号:2019GKTSCX033)
广东省普通高校重点科研项目(编号:2022ZDZX4079)。
关键词
智能驱动
无损动态检测
蜜柚
反馈校正机制
intelligence-driven
nondestructive dynamic testing
pomelo
feedback correction mechanism
