地球信息科学学报 ›› 2013, Vol. 15 ›› Issue (3): 446-451.doi: 10.3724/SP.J.1047.2013.00446

• 遥感科学与应用技术 • 上一篇    下一篇

多时相TM影像决策树模型的水稻识别提取

朱良1,2, 平博2,3, 苏奋振2, 杜云艳2, 苏伟光2   

  1. 1. 兰州交通大学测绘与地理信息学院,兰州730070;
    2. 中国科学院地理科学与资源研究所资源与环境信息系统国家重点实验室,北京100101;
    3. 武汉大学遥感信息工程学院,武汉430079;
    4. 中国科学院烟台海岸带研究所,烟台264003
  • 收稿日期:2012-12-26 修回日期:2013-03-11 出版日期:2013-06-25 发布日期:2013-06-17
  • 通讯作者: 苏奋振(1972-),男,研究员,博士生导师,研究方向为时空信息提取和海洋海岸带应用。E-mail:sfz@lreis.ac.cn E-mail:sfz@lreis.ac.cn
  • 作者简介:朱良(1985-),男,硕士生,湖北襄阳人,研究方向为海岸带遥感信息提取。E-mail:zhul@lreis.ac.cn
  • 基金资助:

    国家自然科学基金项目(41271409)。

Using Decision Tree Model to Extract Paddy Rice Information from Multi-temporal TM Images

ZHU Liang1,2, PING Bo2,3, SU Fenzhen2, DU Yunyan2, SU Weiguang2   

  1. 1. Faculty of Geomatics, Lanzhou Jiaotong University, Lanzhou 730070, China;
    2. State Key Laboratory of Resources and Environment Information System, Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China;
    3. School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 7430079, China;
    4. Yantai Institute of Coastal Zone Research,CAS,Yantai 264003,China
  • Received:2012-12-26 Revised:2013-03-11 Online:2013-06-25 Published:2013-06-17

摘要:

决策树模型的多时相TM影像的小尺度水稻信息提取在我国还鲜有研究。为此,本文利用水稻生长在潮湿土壤这一特性,选取TM影像中对植物含水量和土壤湿度反应敏感的短波红外波段(1.55~1.75μm),以及反映植物覆盖率、植物长势的红光波段(0.62~0.69μm)和近红外波段(0.76~0.96μm),计算水稻移栽期、灌浆期和成熟期3 个时期的归一化植被指数(NDVI)和土壤含水量指数(LSWI),提出一种时间差异的决策树水稻提取模型,以唐山市滦南县南部区域为例开展了研究。经过野外实地验证表明:该模型能有效区分出水域、玉米和菜地等较易与水稻混淆的地物,水稻提取的生产者精度和用户精度分别为95.18%和98.84%,分别比单一时相高出6.78%和7.54%。

关键词: TM影像, 决策树, 信息提取, 水稻, 多时相影像

Abstract:

Remote sensing images have been widely used in extracting and studying paddy rice information and its spatial distribution, which is highly important in land cover analysis, grain structural adjustment and prices making. But studies using multi-temporal TM images to extract small-scale rice information were few. Based on that rice favors growing on wet land, this paper selected three bands of TM images, the short-wave infrared band (1.55-1.75um) which can reflect plant water content and soil moisture, the red band (0.62-0.69um) and the near infrared band (0.76-0.96um), both of which can reflect vegetation coverage and growing condition, to compute the NDVI and LSWI of rice paddles in the three different periods: transplanting stage, heading stage, and maturing stage. Given the two indices demonstrate different characteristics during rice's different growing stages, the paper developed a corresponding time-series-based decision tree model for rice information extraction. A case study using this model was performed in the southern Luannan County, Tangshan City. Been field validated, the experiment results showed the effectiveness of the decision model in distinguishing rice paddles from water area, corn land, farm land and other similar land features, and the producer accuracy and the user accuracy are 95.18% and 98.84%, which outperform the single-temporal results by 6.78% and 7.54% respectively.

Key words: information extraction, multi-temporal images, TM images, paddy rice, decision tree model