摘要
果穗是籽粒的聚集形态。为实现轻量化卷积神经网络对玉米果穗品质的准确、快速识别,提出了一种结合轻量化主干和轻量化通道池化注意力模块(Lightweight channel pooling attention,LCPA)的玉米果穗品质识别模型LCPA-Ghost。首先,采用Ghost网络实现轻量化处理,减少训练成本和冗余信息,提升模型的特征学习能力。其次,将LCPA模块增加到Ghost模块的捷径连接中,在引入少量参数的情况下,弥补空间信息捕获能力的不足,保证模型识别准确率。实验以正常、籽粒杂乱、霉变、杂色和缺粒果穗为研究对象,采集并制作了包含1571张果穗图像的基础数据集。实验结果表明,LCPA-Ghost模型的测试识别率达98.12%,与CorNet相当,而模型参数量仅为2.40 M,单张识别速度为19.08 ms,提升9.8%。LCPA-Ghost模型为玉米果穗品质的轻量化识别提供了可行的实验方法。
Cob,is the aggregated form of the seeds.In order to realize the accurate and fast recognition of corn cob quality by lightweight convolutional neural network,a corn cob quality recognition model LCPA-Ghost combining lightweight backbone and lightweight channel pooling attention was proposed.Firstly,the Ghost network was used to achieve lightweight processing,reduce training cost and redundant information,and improve the feature learning ability of the model.Secondly,the LCPA module was added to the shortcut connection of Ghost module to make up for the lack of spatial information capture capabilities and ensure the model recognition accuracy by introducing a few parameters.The experiments were conducted with normal,seed disorder,mildew,miscellaneous color and missing grain ears,and a base dataset containing 1571 images of cobs was collected and produced.The experimental results indicated that the test recognition rate of LCPA-Ghost model reached 98.12%,comparable to CorNet,while the number of model parameters was only 2.40 M,and the single recognition speed was 19.08 ms,with an improvement of 9.8%.The LCPA-Ghost model provided a feasible experimental method for the lightweight identification of maize ear quality.
作者
刘国荣
史本政
陈召远
徐岩
Liu Guorong;Shi Benzheng;Chen Zhaoyuan;Xu Yan(Shandong University of Science and Technology,Qingdao 266590;Shandong Mobile Communications Group Yantai Branch,Yantai 264000)
出处
《中国粮油学报》
CAS
CSCD
北大核心
2024年第5期165-171,共7页
Journal of the Chinese Cereals and Oils Association
基金
山东省研究生教育优质课程项目(SDYKC19083)
山东省研究生教育联合培养基地项目(SDYJD18027)
海信冰箱公司项目(HSDU20221016)。
作者简介
第一作者:刘国荣,男,2000年出生,硕士,图像处理,计算机视觉,lgr13370891682@126.com;通信作者:徐岩,男,1970年出生,教授,计算机视觉,图像识别与信号处理,xuyan@sdust.edu.cn。
