| 摘要: |
| ［目的］尝试将作物种植结构提取结果由行政单元发展为基于相对均质的地理网格单元，解决当前农作物种植结构信息提取的空间局限性问题，文章在大尺度的土地利用/覆被分类与地块尺度的作物分类之间提出了作物种植结构单元概念，并构建了一种快速、低成本、准确的区域尺度作物种植结构提取方法。［方法］利用黑龙江省2014年250m分辨率的植被指数产品构建时间序列曲线提取物候信息，在耕地物候分区基础上对各物候区进行面向对象的多尺度分割，提取作物种植结构单元，利用光谱特征和NDVI指数构建特征空间，最终采用最邻近分类方法提取作物种植结构。［结果］（1）利用MODIS时间序列数据提取物候特征进行多尺度分割的方法，能够有效的提取区域尺度农作物种植结构单元； （2）作物种植结构提取总体精度为95.70%； （3）黑龙江省2014年作物种植类型共有12种。其中，三江平原主要是水稻单一种植区、水稻混作区； 松嫩平原以玉米单一种植区以及玉米-大豆混作区种植为主； 西北部种植结构较复杂； 东南部因地势等影响多种植玉米、大豆。［结论］利用物候数据进行种植结构提取可以有效划分农业区划，研究成果不仅为作物种植结构调整和农业发展布局提供科学依据，也是不同区域产量预测的基础，为合理布局农业生产、改进耕作制度以及引入和推广新产品等提供依据。 |
| 关键词: 作物种植结构 面向对象分割 多尺度分割 农业区划 时间序列 |
| DOI：10.7621/cjarrp.1005-9121.20170807 |
| 分类号: |
| 基金项目:国家自然科学基金项目“基于土壤不同理化参数光谱特征差异的黑土有机质高光谱遥感反演研究”（40801167）； 黑龙江省普通高等学校新世纪优秀人才培养计划“松嫩平原北部主要土壤有机质光谱速测研究”（1254-NCET-002）； 黑龙江省自然科学基金“区域土壤有机质野外高光谱遥感预测模型研究”（D201404） |
|
| REMOTE SENSING EXTRACTION OF CROP PLANTING STRUCTURE ORIENTED TO AGRICULTURAL REGIONALIZATION |
|
Liu Huanjun, Yan Yan, Zhang Xinle, Qiu Zhengchao, Wang Nan, Yu Wei
|
|
College of Resources and Environmental Sciences, Northeast Agricultural University, Harbin, Heilongjiang 150030, China
|
| Abstract: |
| The aims are to develop the acquisition of crop planting structure information from the administrative unit to the relatively homogeneous geographic grid unit, and solve the problem of space limitations. At present, studies on the acquisition of planting structure information are mainly about the classification of crops at plot scale, or the classification of land use at large scale. Studies on extracting the crop planting structure were too few and limited the application of MODIS image in crop classification. This paper extracted the crop planting structure at regional scale using the remote sensing datasets of 16d and 250m MOD13Q1 time series in 2014, crop phenology parameters, crop phenology parameters in Heilongjiang province according to the curve variation characteristics, partitioned the farmland on the basis of extracting phonological information, and extracted the multi scale segmentation of each phonological region in Heilongjiang province. Finally, the crop planting structure in Heilongjiang Province was extracted by supervised classification. The results showed as follows: （1) The method proposed in this study can effectively extract crop planting structure information at regional scale. （2) The planting structure was verified by using the sample plots from agricultural insurance company, the average accuracy of each type of crop planting structure was 95.70%; （3) The main crops in Heilongjiang province were corn, rice, soybean, potato and wheat12 crops, of which rice single cropping area and rice mixed intercropping area were the main planting structure in Sanjiang plain were, and corn single cropping area and corn-soybean mixed intercropping area were the main structure in Songnen plain. The planting structure was very complicated due to the abundant types of crops in the northwest, the terrain and other factors. The results not only provided a scientific basis for the policy formulation of grain production and agricultural development layout, but also became the foundation of yield prediction in different region, which had great significance to the sustainable development of agriculture. |
| Key words: scrop planting structure object-oriented segmentation multi-scale segmentation agriculture regionalization time series |
