基于YOLO模型的小麦外观分类算法研究  

Research on Wheat Appearance Classification Algorithm Based on YOLO Model

作　　者：徐佳鹏 张朝晖[1,2] 李智 左增杨[1] 赖新亮 赵小燕 张天尧 尹玉国[5] XU Jiapeng;ZHANG Zhaohui;LI Zhi;ZUO Zengyang;LAI Xinliang;ZHAO Xiaoyan;ZHANG Tianyao;YIN Yuguo(School of Automation and Electrical Engineering,University of Science and Technology Beijing,Beijing 100083,China;Beijing Engineering Research Center of Industrial Spectrum Imaging,Beijing 100083,China;College of Information Science and Technology,Henan University of Technology,Zhengzhou 450001,China;Henan Provincial Key Laboratory of Grain Photoelectric Detection and Control,Zhengzhou 450001,China;Shandong Start Measurement and Control Equipment Co.,Ltd.,Weifang 261041,China)

机构地区：[1]北京科技大学自动化学院,北京100083 [2]北京市工业波谱成像工程技术研究中心,北京100083 [3]河南工业大学信息科学与工程学院,河南郑州450001 [4]河南省粮食光电探测与控制重点实验室,河南郑州450001 [5]山东思达特测控设备有限公司,山东潍坊261041

出　　处：《自动化仪表》2023年第3期83-87,共5页Process Automation Instrumentation

基　　金：国家重点研发计划基金资助项目(2019YFB2101902);中央高校基本科研业务基金资助项目(FRF-TP-20-015A1)。

摘　　要：小麦种植广泛且营养丰富,其品质问题需要重点关注。小麦品质的衡量指标主要是不完善粒占比。为此,需要对小麦颗粒进行分类识别。提出了1种基于你只看一次(YOLO)模型的小麦外观自动分类算法,创新性地将YOLO模型应用于小麦外观分类场景。对采集得到的小麦样本图像切割、筛选、扩充和标记,构建了完善粒与不完善粒图像库。对YOLO网络进行了训练,利用训练后的模型对麦粒图像进行了测试,实现了完善粒、不完善粒分别为91.7%、87.1%的分类准确率。这种自动分拣麦粒的检验方法避免了人工视觉疲劳后的误判,而且检测效率显著提高,为小麦外观分类研究提供了新的思路。Wheat is widely grown and nutrient-rich,and its quality issues require focused attention.The main measure of wheat quality is the percentage of imperfect grains.For this reason,wheat grains need to be classified and identified.An automatic wheat appearance classification algorithm based on the you only look once(YOLO)model is proposed,and the YOLO model is innovatively applied to the wheat appearance classification scenario.The collected wheat sample images are cut,filtered,expanded,and labeled to construct a library of perfect and imperfect grains images.The YOLO network is then trained,and the wheat grain images are tested using the trained model,achieving classification accuracies of 91.7%and 87.1%for perfect and imperfect grains,respectively.This test method for automatic sorting of wheat grains avoids misclassification after manual visual fatigue,and the detection efficiency is significantly improved,providing a new idea for wheat appearance classification research.

关 键 词：小麦 不完善粒 外观分类 图像检测 图像识别 深度学习 你只看一次模型 卷积神经网络 

分 类 号：TH79[机械工程—仪器科学与技术]