基于深度学习Caffe框架的瓜蒌子完整度检测方法  

A Completeness Detection Method for Melon Seeds Based on Deep Learning Caffe Framework

作　　者：刘亚伟 易克传 张前景 王川[2] Liu Yawei;Yi Kechuan;Zhang Qianjing;Wang Chuan(College of Mechanical Engineering,Anhui Science and Technology University,Fengyang,Anhui 233100,China;Agriculture Mechanization and Engineering Research Institute,Anhui Academy of Agricultural Sciences,Hefei,Anhui 230001,China)

机构地区：[1]安徽科技学院机械工程学院,安徽凤阳233100 [2]安徽省农业科学院农业机械装备与工程研究所,安徽合肥230001

出　　处：《黑龙江工业学院学报（综合版）》2024年第8期91-95,共5页Journal of Heilongjiang University of Technology(Comprehensive Edition)

基　　金：安徽省重点研发计划项目“高效瓜蒌取籽清选设备的研发及应用”(项目编号:202204c06020008)。

摘　　要：为了进一步发展农业方向的自动化、无人化,改善在农产品检测方面使用人工检测的速度慢、精度低、费时费力的现状,为了降低图像处理中计算所使用的数据量,同时也为了提高图像信息的易读性,将构建瓜蒌子图像采集系统获取与环境颜色对比度明显的待检测瓜蒌子图像。利用深度学习Caffe框架中的深度卷积神经网络模型对瓜篓子完整度进行综合识别,使用了深度学习Caffe架构中的深度卷积式神经网络模型对瓜篓子的完整程度进行了综合辨识。结果得到瓜篓子整齐程度的综合识别率,再与支持向量机语言(SVM)的加以对比。深度学习Caffe架构下的深度卷积神经网络模式的识别率,相对于支持向量机器(SVM)的方式进行了提升,且成效更加突出。In order to further develop the automation and unmanned agricultural direction,improve the current situation of slow speed,low accuracy,time-consuming and laborious use of manual detection in farm product inspection,reduce the amount of data used in image processing calculation,and also improve the legibility of image information,the melon seed image acquisition system will be built to obtain the image of the melon seeds to be detected with obvious color contrast with the environment.A deep convolutional neural network model in the Deep Learning Caffe framework was used for comprehensive recognition of the completeness of the melon basket.The combined recognition rate of the degree of neatness of the melon basket is obtained and then compared with that of the Support Vector Machine Language(SVM).The recognition rate of deep convolutional neural network patterns under the deep learning Caffe architecture is improved and more effective relative to the more support vector machine(SVM)approach.

关 键 词：深度学习 卷积神经网络 瓜蒌子完整度 Caffe框架 支持向量机 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]