Crop Classification Based on Time Series MODIS EVI and Ground Observation for Three Adjoining Years in Xinjiang
Shakir Muhammad1,2, NIU Zheng1*, WANG Li1,2, Abdullah Aablikim3, HAO Peng-yu1,2, WANG Chang-yao1
1. State Key Laboratory of Remote Sensing Science, Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100101, China 2. University of Chinese Academy of Sciences, Beijing 100049, China 3. National Bureau of Statistic China Survey Office in Xinjiang, Urumqi 830001
摘要： 在区域或全球尺度，250 m分辨率的MODIS EVI常被用于作物分类。而且，基于遥感数据可以快速准确的进行作物分类，并为辅助农业政策的制定，因而得到了广大研究者的关注。研究提出了直接使用多年MODIS 250 m EVI和临近年份地面调查数据进行作物分类的方法。首先，通过扩展2011，2012和2013年的野外调查数据获得全疆的典型地块，并从地块中提取MODIS纯像元作为分类样本。接着使用免疫系统网络分类器(ABNet)提取研究取的主要作物，包括棉花、玉米、冬小麦和葡萄等。在三年的数据中，任意两年的地面数据用于训练分类器，用使用训练好的分类器对另一年的数据进行分类。例如，使用2011和2012年的数据训练分类器，并对2013年的数据进行分类。结果表明，每年的分类精度达80%以上，且Kappa系数高于0.7。今后工作中，仍需使用更多的地面数据获得更的更精细的分类结果。
关键词：MODIS;增强植被指数(EVI);免疫系统网络(ABNet);土地覆盖分类
Abstract：There is a regular use of Moderate Resolution Imaging Spectroradiometer (MODIS) 250 meter EVI to classify the crops on a regional level throughout the world. A rapid agricultural land use change attributed to new Chinese agriculture policy is attracting many researchers to focus. The objective of this study is to present a more straightforward multiyear classification methodology using time series MODIS EVI with 250 meters spatial resolution and subsequent field data in Xinjiang, China. An extensive polygon based ground reference annual crop data were collected for the years 2011, 2012 and 2013 throughout the study area. The most pure pixel within each polygon was selected which eases crop differentiation. Artificial Immune Network (ABNet) was used to classify cotton, maize, wheat/others, rice and grapes, dominating most of the study area. The data of two different years were used together to classify the crop of next year, as 2011 and 2012 were used to classify crops of 2013. Classification results were validated using the same year ground data. Results showed the classification accuracy above 80% for each year with kappa coefficient of 0.7 and above. However more research and additional ground reference data are needed to classify a range of crops in the study area which will give a more detailed view of the land use land cover change strengthening agriculture decisions practices in the future.
Shakir Muhammad1,2，牛 铮1*，王 力1，Abdullah Aalikim3，郝鹏宇1,2，王长耀1 . 基于多时相MODIS EVI和临近三年地面数据的新疆作物分类 [J]. 光谱学与光谱分析, 2015, 35(05): 1345-1350.
Shakir Muhammad1,2, NIU Zheng1*, WANG Li1,2, Abdullah Aablikim3, HAO Peng-yu1,2, WANG Chang-yao1. Crop Classification Based on Time Series MODIS EVI and Ground Observation for Three Adjoining Years in Xinjiang. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2015, 35(05): 1345-1350.
[1] Clark M L, Aide T M, Grau H R, et al. Remote Sensing of Environment, 2010, 114(11): 2816. [2] DeFries R S, Morton D C, van der Werf G R, et al. Geophysical Research Letters, 2008, 35(22): L22705. [3] Brown J, Jepson W, Kastens J, et al. GI Science & Remote Sensing, 2007, 44(2): 117. [4] Brown J C, Kastens J H, Coutinho A C, et al. Remote Sensing of Environment, 2013, 130: 39. [5] Wardlow B, Egbert S, Kastens J. Remote Sensing of Environment, 2007, 108(3): 290. [6] Homer C, Huang C, Yang L, et al. Photogrammetric Engineering & Remote Sensing, 2004, 70(7): 829. [7] Lu H, Raupach M R, McVicar T R, et al. Remote Sensing of Environment, 2003, 86(1): 1. [8] Sakamoto T, Wardlow B D, Gitelson A A, et al. Remote Sensing of Environment, 2010, 114(10): 2146. [9] Wardlow B D, Egbert S L. Remote Sensing of Environment, 2008, 112(3): 1096. [10] Yanfei Z, Liangpei Z. Geoscience and Remote Sensing, IEEE Transactions on, 2012, 50(3): 894. [11] Xie Y, Sha Z, Yu M. Journal of Plant Ecology, 2008, 1(1): 9. [12] Mingwei Z, Qingbo Z, Zhongxin C, et al. International Journal of Applied Earth Observation and Geoinformation, 2008, 10(4): 476.
