基于遥感与多变量概率抽样调查的作物种植面积测量

Mapping crops acreages based on remote sensing and sampling investigation by multivariate probability proportional to size

针对传统抽样调查工作中调查基础资料时效性不高和野外调查工作量较大等问题,该文提出了一种遥感与MPPS(multivariate probability proportional to size)抽样调查相结合的农作物种植面积测量方法。利用第2次农业普查数据进行抽样框的编制;利用时序中分辨率遥感数据进行农作物种植面积的分类;在中分辨率遥感分类图的基础上进行MPPS抽样;采用高空间分辨率遥感数据对抽选样本进行面向对象的分类;根据MPPS抽样方法进行总体农作物种植面积的推断;计算CV值,评价抽样精度,以国家统计局公布数据为标准进行总体面积精度评价。以辽宁省北镇市为研究区对该方法进行了测试。结果显示,该方法能够有效的提取县级农作物种植面积,农作物种植面积提取精度优于92%。

MPPS is a method widely used in crop area statistics in the Chinese crop area statistical investigation business. However, this method has two drawbacks. One is the outdated basic data. The other is the large workload of a field survey. The second land use survey data used as the basic data in the Chinese crop area statistical investigation is only updated every 10 years. The longer update cycle makes it difficult to react to the inter-annual change of crop areas. The artificial field survey is used in the Chinese crop area statistical investigations to survey the area of crops of every sampling village. Because of the large number of sample villages, the workload of field investigation is huge, and time-consuming and laborious. In order to solve those problems in a conditional sampling survey, a novel crop area extraction method was proposed in this paper using remote sensing and MPPS sampling technology. The sampling frame was prepared using the village-level administrative unit data of the second land use survey data. Crops were extracted using multi-temporal HJ-1 satellite data with a Spectral Angle Mapper method. Three HJ-1 satellite data sets acquired in April, May, and August were selected according to the Phonological data. In April, rice and winter wheat were in the seedling stage, and corn was not planted. In May, rice was in irrigation period. In August, winter wheat had been harvested, while rice and corn were in their maturity periods. So using images in those months, it was easy to differentiate rice from winter wheat and corn since the paddy land contains water, while the wheat and corn land were dry in May. It was also easy to differentiate winter wheat from corn because the growing period of winter wheat was 20 days earlier than the growing period of corn. Then the crop resolution crop classification map. Combining the updated areas of each village were updated by the moderate sampling frame data and MPPS sampling method, sampling villages were selected. Crops in the sampling villages were mapped using ZY-1 02c satellite data with an object-oriented classification method. The ZY-1 02C satellite is a new Chinese civil remote sensing satellite launched on December 22, 2011. It was the highest resolution civil remote sensing satellite in China which carried a panchromatic/multispectral sensor and a high-resolution sensor. The spatial resolution of the HR sensor was 2.36 m, and the spatial resolution of the panchromatic/multispectral sensor was 5m in a panchromatic band and 10m in three multispectral bands. Finally, according to the MPPS method, the total area of each crop in the study area and CV were calculated. The algorithm had been tested over a study area in Beizhen Country, Liaoning Province, China. The results showed that this method could effectively determine the rice and corn areas. A high mapping precision of 92% was obtained.