地球信息科学学报 ›› 2019, Vol. 21 ›› Issue (5): 789-798.doi: 10.12082/dqxxkx.2019.180418

• 遥感科学与应用技术 • 上一篇    

无人机影像的面向对象水稻种植面积快速提取

吴方明(), 张淼, 吴炳方*()   

  1. 中国科学院遥感与数字地球研究所 遥感科学国家重点实验室,北京,100101
  • 收稿日期:2018-08-29 修回日期:2019-02-20 出版日期:2019-05-25 发布日期:2019-05-25
  • 通讯作者: 吴炳方 E-mail:wufm@radi.ac.cn;wubf@radi.ac.cn
  • 作者简介:

    作者简介:吴方明(1984-),男,湖北汉川人,硕士,工程师,主要从事无人机遥感研究。E-mail: wufm@radi.ac.cn

  • 基金资助:
    中国科学院科技服务网络计划(STS计划)项目(KFJ-STS-ZDTP-009);国家自然科学基金项目(41561144013、41861144019、41701496)

Object-oriented Rapid Estimation of Rice Acreage from UAV Imagery

Fangming WU(), Miao ZHANG, Bingfang WU*()   

  1. State Key Laboratory of Remote Sensing Science, Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100101, China
  • Received:2018-08-29 Revised:2019-02-20 Online:2019-05-25 Published:2019-05-25
  • Contact: Bingfang WU E-mail:wufm@radi.ac.cn;wubf@radi.ac.cn
  • Supported by:
    Science and Technology Service Network Initiative (STS), No.KFJ-STS-ZDTP-009;National Natural Science Foundation of China, No. 41561144013, 41861144019, 41701496.

摘要:

基于抽样技术的地面调查与遥感影像分类相结合的方法在大范围作物种植面积提取中得到广泛使用。无人机影像具有低成本、高时效、高分辨率的一系列优点,可以快速实现特定区域范围内的农情采样任务。本文以水稻样地为研究对象,采用便携式无人机Mavic Pro进行航拍。对所获取无人机影像进行预处理生成分辨率为3.95cm/pix的正射影像,采用面向对象的思想,目视评价和ESP工具相结合快速选择了最优分割尺度为300,应用了支持向量机、随机森林和最邻近监督分类方法对影像进行了地物分类和水稻面积快速提取。采用目视解译分类结果进行分类结果和面积精度评价,总体精度最高的方法为最邻近分类法,此时水稻分类用户精度为95%,面积一致性精度为99%。研究结果说明了无人机遥感和自动分类能够在平原水稻种植区快速获取样方内高分辨率影像并提取水稻种植面积,弥补了农田被遮挡时地面调查数据的缺失,为大范围水稻种植面积、产量等信息的计算提供样本和验证依据。

关键词: 无人机影像, 水稻, 面向对象分割, 最近邻监督分类, 种植面积

Abstract:

The methodology of combining sampling-based ground survey and satellite imagery classification has been widely used in estimating crop acreage on large scales. Use of unmanned aerial vehicle (UAV) imagery has a series of merits including low cost, high efficiency, and high resolution, which make it possible to quickly monitor the agricultural conditions over a specific area. With a research focus on rice sample plots, this study used a portable UAV Mavic Pro to obtain aerial imagery. The UAV imagery were preprocessed to generate an orthophoto with a resolution of 3.95 cm/pix. By adopting the object-oriented classification philosophy, visual assessment, and the Estimation of Scale Parameter (ESP) tool, the optimal segmentation scale was determined to be 300. The support vector machine, random forest, and nearest neighborhood classifiers were employed and contrasted for imagery classification and the extraction of rice acreage; visual interpretation was used for assessing the accuracy of the classification results. The best automatic classification method turned out to be nearest neighborhood classification, with its user accuracy of rice being 95% and the area consistency accuracy 99%. The findings show that use of UAV imagery and automatic classification can quickly acquire high-resolution imagery and extract rice acreage in rice growing areas on plains. Moreover, high-resolution UAV imagery can be used as ground truth data when cropland is in shadow. The proposed approach helps provide validation samples for estimating rice acreage and production on large scales.

Key words: UAV imagery, rice, object-oriented segmentation, nearest neighbor supervised classification, planted area