摘要
为解决青椒智能识别问题,以在自然环境中采集的苏椒1614图像为识别对象,采用深度学习方法,选择YOLO-v3,Faster R-CNN和CenterNet 3种神经网络进行深度学习模型训练,并比较分析不同深度学习模型的识别结果。试验结果表明,Faster R-CNN为青椒识别的最优模型,其精度、召回率和F1值分别达到92.4%,79%和85.2%,证明深度学习方法能够有效提取图像特征。研究为青椒的智能化识别与采摘提供依据。
In order to solve the problem of intelligent recognition of green peppers,1614 images of Su-pepper collected in natural environment were used as the recognition object,deep learning method was adopted,and three neural networks of YOLO-v3,Faster R-CNN and CenterNet were selected for deep learning model training,and the recognition results of different deep learning models were compared and analyzed.The experimental results show that Faster R-CNN is the optimal model for recognition of green peppers,and its accuracy,recall rate and F1 values reach 92.4%,79%and 85.2%,respectively.This study also proves that the deep learning method can effectively extract image features,which provides a basis for intelligent recognition and picking of green peppers.
作者
汪谦谦
孙艳霞
徐星星
金小俊
于佳琳
陈勇
WANG Qianqian;SUN Yanxia;XU Xingxing;JIN Xiaojun;YU Jialin;CHEN Yong(College of Mechanical and Electronic Engineering,Nanjing Forestry University,Nanjing 210037,China;School of Rail Transportation,Nanjing Vocational Institute of Transport Technology,Nanjing 211188,China;Institute of Modern Agriculture,Peking University,Weifang 261000,China;Department of Soil and Crop Sciences,Texas A&M University,College Station,TX 77843,USA)
出处
《包装与食品机械》
CAS
北大核心
2023年第3期89-93,共5页
Packaging and Food Machinery
基金
国家自然科学基金项目(32072498)
江苏省研究生科研与实践创新计划项目(KYCX22_1051)。
关键词
青椒识别
自然环境
深度学习
智能化采摘
green pepper recognition
natural conditions
deep learning
intelligent picking
