欢迎您访问《智慧农业(中英文)》官方网站! English

Smart Agriculture ›› 2022, Vol. 4 ›› Issue (4): 49-60.doi: 10.12133/j.smartag.SA202207004

• 专题--大田作物智慧种植 • 上一篇    下一篇

基于多种深度学习算法的田间玉米籽粒检测与计数

刘晓航1,2(), 张昭1,2(), 刘嘉滢1,2, 张漫1,2, 李寒1,2, FLORES Paulo3, 韩雄哲4,5   

  1. 1.中国农业大学智慧农业系统集成研究教育部重点实验室,北京 100083
    2.中国农业大学农业农村部农业信息获取技术重点实验室,北京 100083
    3.北达科他州州立大学农业与生物工程系,法戈 58102,美国
    4.韩国江原大学生物系统工程系,春川 24341,韩国
    5.韩国江原大学智慧农业交叉学科,春川 24341,韩国
  • 收稿日期:2022-07-12 出版日期:2022-12-30
  • 基金项目:
    中央高校基本科研业务费专项资金资助(00112502);世界顶尖涉农大学国际合作交流种子基金(15052001)
  • 作者简介:刘晓航(1996-),男,博士研究生,研究方向为智慧农业。E-mail:lxhhaust@163.com
  • 通信作者: 张 昭(1985-),男,博士,教授,研究方向为智慧农业。E-mail:zhaozhangcau@cau.edu.cn

Infield Corn Kernel Detection and Counting Based on Multiple Deep Learning Networks

LIU Xiaohang1,2(), ZHANG Zhao1,2(), LIU Jiaying1,2, ZHANG Man1,2, LI Han1,2, FLORES Paulo3, HAN Xiongzhe4,5   

  1. 1.Key Laboratory of Smart Agriculture System Integration, Ministry of Education, China Agricultural University, Beijing 100083, China
    2.Key Lab of Agricultural Information Acquisition Technology, Ministry of Agriculture and Rural Affairs, China Agricultural University, Beijing 100083, China
    3.Department of Agricultural and Biosystems Engineering, North Dakota State University, Fargo, ND 58012, United States
    4.Department of Biosystems Engineering, Kangwon National University, Chuncheon 24341, Korea
    5.Interdisciplinary Program in Smart Agriculture, Kangwon National University, Chuncheon 24341, Korea
  • Received:2022-07-12 Online:2022-12-30
  • Foundation items:Central University Basic Research Business Fee Special Fund Support (00112502); World's Top Agricultural University International Cooperation and Exchange Seed Fund (15052001)
  • About author:LIU Xiaohang, E-mail:lxhhaust@163.com
  • Corresponding author:ZHANG Zhao, E-mail:zhaozhangcau@cau.edu.cn

摘要:

为快速准确获取玉米收获过程中遗失籽粒数信息,进行收割损失调节等管理,对比评估了单阶段和两阶段主流目标检测网络对田间玉米籽粒计数的性能。首先,利用RGB相机获取包含不同背景和不同光照的图像数据,并进一步生成数据集;其次,构建籽粒识别的不同目标检测网络,包括Mask R-CNN、EfficientDet-D5、YOLOv5-L、YOLOX-L,并利用所采集的420幅有效图像对构建的四种网络进行训练、验证、测试,图像数分别为200、40和180幅;最后,依据测试集图像的识别结果进行籽粒计数性能评价。试验结果表明,YOLOv5-L网络对测试集图像检测的平均精度为78.3%,模型尺寸仅为89.3 MB;籽粒计数的检测正确率、漏检率和F1值分别为90.7%、9.3%和91.1%,处理速度为55.55 f/s,识别与计数性能均优于Mask R-CNN、EfficientDet-D5和YOLOX-L网络,并对具有不同地表遮挡程度和籽粒聚集状态的图像具有较强的鲁棒性。深度学习目标检测网络YOLOv5-L可实现实际作业中玉米收获损失籽粒的实时监测,精度高、适用性强。

关键词: 收获损失, 田间玉米籽粒, 深度学习, 籽粒计数, YOLOv5-L, YOLOX-L, Mask R-CNN, EfficientDet-D5

Abstract:

Machine vision has been increasingly used for agricultural sensing tasks. The detection method based on deep learning for infield corn kernels can improve the detection accuracy. In order to obtain the number of lost corn kernels quickly and accurately after the corn harvest, and evaluate the corn harvest combine performance on grain loss, the method of directly using deep learning technology to count corn kernels in the field was developed and evaluated. Firstly, an RGB camera was used to collect image with different backgrounds and illuminations, and the datasets were generated. Secondly, different target detection networks for kernel recognition were constructed, including Mask R-CNN, EfficientDet-D5, YOLOv5-L and YOLOX-L, and the collected 420 effective images were used to train, verify and test each model. The number of images in train, verify and test datasets were 200, 40 and 180, respectively. Finally, the counting performances of different models were evaluated and compared according to the recognition results of test set images. The experimental results showed that among the four models, YOLOv5-L had overall the best performance, and could reliably identify corn kernels under different scenes and light conditions. The average precision (AP) value of the model for the image detection of the test set was 78.3%, and the size of the model was 89.3 MB. The correct rate of kernel count detection in four scenes of non-occlusion, surface mid-level-occlusion, surface severe-occlusion and aggregation were 98.2%, 95.5%, 76.1% and 83.3%, respectively, and F1 values were 94.7%, 93.8%, 82.8% and 87%, respectively. The overall detection correct rate and F1 value of the test set were 90.7% and 91.1%, respectively. The frame rate was 55.55 f/s, and the detection and counting performance were better than Mask R-CNN, EfficientDet-D5 and YOLOX-L networks. The detection accuracy was improved by about 5% compared with the second best performance of Mask R-CNN. With good precision, high throughput, and proven generalization, YOLOv5-L can realize real-time monitoring of corn harvest loss in practical operation.

Key words: harvest loss, infield corn kernel, deep learning, kernel count, YOLOv5-L, YOLOX-L, Mask R-CNN, EfficientDet-D5

中图分类号: