引用本文:安江勇,黎万义,李茂松※.基于Mask R-CNN 的玉米干旱卷曲叶片检测[J].中国农业信息,2019,31(5):66-74
【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 69次   下载 40 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于Mask R-CNN 的玉米干旱卷曲叶片检测
安江勇1, 黎万义2, 李茂松※1
1.中国农业科学院农业资源与农业区划研究所/ 农业农村部农业遥感重点实验室,北京100081;2.中国科学院自动化研究所,北京100190
摘要:
【目的】干旱胁迫严重影响玉米生长和产量水平,对玉米干旱胁迫快速、精确监测,及时制定相应的防灾减灾措施对保障玉米丰产稳产具有重要意义。对玉米干旱卷曲叶片进行监测是实现快速精确地监测玉米干旱胁迫的重要方法。【方法】文章使用单反相机获取干旱胁迫和适宜水分处理下的玉米植株数字图像,使用多边形框手动标注玉米卷曲叶片,建立玉米卷曲叶片目标检测数据集,使用目标检测模型Mask R-CNN 对玉米卷曲叶片进行检测。【结果】目标检测模型进行玉米卷曲叶片检测的置信度高于98%,在IOU 阈值为0.5 时,卷曲叶片检测模型的均值平均精度为74.35%。【结论】目标检测算法能精确地对玉米卷曲叶片进行检测和分割,卷曲检测置信度高。基于叶片卷曲信息的玉米干旱胁迫识别具有快速、及时、精确等优点。随着作物表型组学的发展,目标检测算法可广泛应用于作物生物胁迫、非生物胁迫以及作物表型研究中感兴趣区域的识别和定位等研究。
关键词:  干旱;玉米植株;叶片卷曲;目标检测
DOI:10.12105/j.issn.1672-0423.20190507
分类号:
基金项目:国家科技支撑计划课题“农业干旱与干热风监测预警与应急防控关键技术研究”(2012BAD20B01);中国农业科学院科技创新工程“农业灾害监测预警新技术新方法研发”(CAAS-ASTIP-IARRP-2013)
Identification of maize drought rolled leaves based on Mask R-CNN model
An Jiangyong1, Li Wanyi2, Li Maosong※1
1.Institute of Agricultural Resources and Regional Planning,Chinese Academy of Agricultural Sciences/ Key Laboratory of Agricultural Remote Sensing,Ministry of Agriculture and Rural Affairs,Beijing 100081,China;2.Institute of Automation,Chinese Academy of Sciences,Beijing 100190,China
Abstract:
[ Purpose]Drought stress seriously affects maize growth and yield level. Rapid and accurate monitoring maize drought stress and timely formulation disaster prevention and mitigation measures are of great significance to ensure high and stable maize yield. Detecting the drought rolled leaves of maize is an important method to monitor maize drought stress fast and accurately. [ Method]In this paper,the digital image of maize under drought stress and suitable water treatment was obtained by using SLR camera. The rolled leaves of maize were labeled manually with polygonal frame,and established a rolled leaf object detection dataset. An object detection model Mask R-CNN was applied to detection maize rolled leaves.[Result]The rolled leaves detection confidence of the object detection model was higher than 98%. When IOU threshold was 0.5,the average accuracy of the model was 74.35%.[Conclusion]Object detection model can accurately detect and segment maize rolled leaves,and has high confidence in rolled detection. The identification of maize drought stress based on leaf rolled information is fast,timely and accurate. With the development of crop phenomics,object detection algorithms can be widely used in crop biotic and abiotic stresses as well as identification and location of regions of interest in phenotypic studies.
Key words:  drought;maize plant;rolled leaf;object detection