基于深度学习的烟叶烘烤实时识别研究  被引量:1

Research on Real-time Recognition of Tobacco Curing Based on Deep Learning

在线阅读下载全文

作  者:张成双[1] 王先伟[1] 刘志刚[1] 王桂荣[1] 姜增昀 郝允志[2] ZHANG Chengshuang;WANG Xianwei;LIU Zhigang

机构地区:[1]山东潍坊烟草有限公司,山东潍坊261000 [2]西南大学,重庆400715

出  处:《智慧农业导刊》2022年第21期18-22,共5页JOURNAL OF SMART AGRICULTURE

基  金:山东潍坊烟草有限公司科技项目(潍烟技[2021]2号)。

摘  要:烟叶烘烤阶段智能识别是实现烟叶智能烘烤至关重要的一步,通过烟叶实时状态科学确定烘烤阶段是实现烟叶精准烘烤的必然要求。目前密集烤房烘烤主要通过烘烤师主观设定烘烤曲线的方式来确定烘烤工艺,难以保证烘烤工艺的实时精准匹配,为提高烟叶烘烤质量,减轻烘烤劳动负担,该文通过烟叶智能烘烤仪采集大量现场图片数据,并通过双边滤波方式进行去噪,创建烘烤阶段数据集。提出一种采用高效ECA注意力模块的Mobile Netv3-ECA模型对图像数据进行识别。对训练参数进行优化调整后,在Mobile Netv3-ECA网络上准确率达到91.38%,优化后的模型提高烟叶烘烤阶段控制的精准性,可以根据烟叶实际状态推算合适的烘烤工艺,并提高烟叶烘烤工艺的精准度,对提高烟农收入、促进区域经济发展具有实际意义。Intelligent identification of tobacco curing stage is an important step to realize intelligent tobacco curing,and scientific determination of tobacco curing stage through the real-time state of tobacco leaves is an inevitable requirement to achieve accurate tobacco curing.At present,the curing process of dense curing room is mainly determined by the baking curve set subjectively by the baker,so it is difficult to ensure the real-time accurate matching of the curing process.In order to improve the tobacco curing quality and reduce the baking burden,this paper collects a large number of on-site picture data through the tobacco intelligent curing instrument,and denoises by bilateral filtering,and creates the data set of the curing stage.A MobileNetv3-ECA model based on efficient ECA attention module is proposed to recognize image data.After optimizing and adjusting the training parameters,the accuracy on the MobileNetv3-ECA network reaches 91.38%.The optimized model improves the accuracy of tobacco curing stage control,helps calculate the appropriate curing process according to the actual state of tobacco leaves,improves the accuracy of tobacco curing process,and thus has practical significance to improve the income of tobacco farmers and promote regional economic development.

关 键 词:烟叶烘烤 双边滤波 深度学习 神经网络 烘烤阶段 

分 类 号:S-3[农业科学]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象