尺度变化对油菜种植面积遥感监测精度的影响

INFLUENCE OF SCALE VARIATION ON PRECISION OF RAPESEED ACREAGE REMOTE SENSING MONITORING

关键词:  尺度变化 油菜面积 遥感监测 精度

Key words:  scale variation  rape acreage  remote sensing  monitoring  precision

摘要:

农作物种植面积监测是遥感估产的基础工作之一，监测结果的精度直接影响到数据的可信度和可用性。随着多光谱和多空间分辨率数据集的日益普遍，多尺度遥感数据源的综合应用是进行大范围农作物种植面积遥感监测业务化运行的必然趋势，因此尺度变化对农作物种植面积遥感监测精度影响的研究势在必行。该研究采用高空间分辨率的多光谱遥感影像进行油菜种植面积提取，对其提取结果进行基于简单多数原则的尺度转化，得到不同空间分辨率的提取结果，通过与地面实测样方数据构建误差矩阵进行精度分析，分析不同空间分辨率影像对作物种植面积遥感信息提取精度的影响。运用多种精度评价指标，从不同空间分辨率、不同监测方法、不同遥感监测时相等角度系统分析了农作物种植面积遥感测量中的尺度效应问题。为基于多尺度遥感数据综合应用的农作物种植面积遥感监测业务化运行中的数据选择和精度保证问题提供理论与实验基础。

Abstract:

The crop acreage monitoring using the remote sensing techniques is one of the basic task in the remote-sensing yield estimation and its precise result would affect the data in reliability and applicability directly. With the development of data sets of multispectral and multi-spatial resolutions, the comprehensive application for multi-scale remote sensing data sources is an evitable trend in the operation of the remote sensing monitoring of the large crop acreage. Thus, it is imperative to study the influence of the scale variation on remote sensing monitoring precision of the crop acreage. Taking Anhui province as the research area and rapeseed as target crop, the cultivated area was extracted using high spatial resolution spectrum of remote sensing image, then the extracted results in different spatial resolutions were available by the scale conversion of the results based on the principle of a simple majority. These results, together with the sample data measured on the ground were used to build up error matrix for analysis of precisions. Meanwhile, the influence of different spatial resolutions on crop acreage monitoring could be analyzed precisely. Remote sensing information sources were tested by many ways of auto information extraction such as non-supervision classification and Mahalanobis Distance Method, including different classification units, different algorithms, and different corrections. The target crop acreages in different compound modes were compared. Meanwhile, the extracted results were evaluated in combination of random distribution points and field sample testing respectively. The remote sensing images（Landsat-TM）in different time sections were collected for covered target crops during main growth periods, and the crop acreages were extracted by the non-supervision-artificial clustering method. The scale effects in crop acreage remote sensing monitoring were analyzed systematically in different spatial resolutions, different monitoring methods and different remote sensing times and many precision evaluation indexes. The results provided the foundation in theory and experiment for data selection and precision guarantee in the crop acreage remote sensing monitoring based on multi-scale remote sensing data．