引用本文:郭涛,郭家,李宗南,邱霞,王思.基于Darknet深度学习框架的桃花检测方法[J].中国农业信息,2021,33(6):25-33
【打印本页】   【HTML】   【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 404次   下载 187 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于Darknet深度学习框架的桃花检测方法
郭涛1,郭家2,李宗南1,邱霞1,王思1
1.四川省农业科学院遥感与数字农业研究所,成都 610066;2.福州大学,数字中国研究院(福建), 福建福州 350108
摘要:
【目的】 为实现果园自然场景下智能农业机器人对桃花的准确、快速、有效检测。【方法】 文章采用相机获取桃花图片数据,通过LabelImg软件进行人工标记建立桃花目标识别的检测样本数据集,训练Darknet深度学习框架下的YOLO v4模型对桃花进行识别。【结果】 模型精度评估表明,YOLO v4模型的平均准确率MAP值(86%)比Faster R-CNN的MAP值(51%)高出35%。【结论】 YOLO v4与经典的算法相比,对各种自然环境下的桃花检测具有较好的实时性和鲁棒性,可为精准识别桃花提供重要参考价值,桃花精准识别为疏花疏果作业奠定了基础。
关键词:  Darknet  Faster R-CNN  桃花识别  目标检测  自然场景
DOI:10.12105/j.issn.1672-0423.20210603
分类号:
基金项目:四川省科技计划项目“农业大数据资产管理及智能分析应用系统”(2021YFG0028)
The peach blossom detection method based on darknet deep learning framework
Guo Tao1, Guo Jia2, Li Zongnan1, Qiu Xia1, Wang Si1
1.Institute of Remote Sensing and Digital Agriculture,Sichuan Academy of Agricultural Sciences,Chengdu 610066,China;2.The Academy of Digital China(Fujian),Fu Jian Fuzhou 350116,China
Abstract:
[Purpose] The purpose is realize the accurate,rapid and effective detection of peach blossom by intelligent agricultural robot in natural orchard scene.[Method] the camera is used to get peach blossom’s image data,and LabelImg software is used for manual marking to establish a detection sample dataset for peach blossom target identification,and YOLO v4 model under Darknet deep learning framework is trained to identify peach blossom.[Result] The accuracy evaluation of the model showed that the MAP accuracy of YOLO v4 model is 86%,35% higher than that of Faster R-CNN,which was 51%.[Conclusion] Compared with the traditional algorithm,the YOLO v4 algorithm has better real-time performance and robustness for peach blossom detection in various natural environments,which has important reference value for accurate peach blossom recognition identification which further lays a foundation for accurate peach blossom and fruit thinning.
Key words:  Darknet  Faster R-CNN  peach blossom identification  target detection  natural scene