基于NDVI时间序列影像的农业种植结构提取——以河北省邯郸市广平县为例

Extraction of Agricultural Planting Structure Based on NDVI Time Series Image—A Case Study of Guangping County, Handan City, Hebei Province

摘要: 农业种植结构提取是基础性环节，针对原来以人工传统调查统计的农作物信息获取方法主观性较强且误差巨大难以满足信息时代背景下对农业的管理。近年来，水资源匮乏不断制约着灌溉农业发展，不同作物在全生育期所需的灌溉量存在巨大差异，因此快速准确的了解种植结构对于节水农业有巨大效益。本文以广平县为研究区，以哨兵2号遥感影像作为基础数据，以人工采集和目视解译的样本点为标准，采用随机森林算法对广平县的种植结构进行分类。结果表明：分类结果的总体精度为89.95%，Kappa系数为0.88，总体精度和Kappa系数均高于0.85，说明各个农作物分类结果较为精准。经统计，广平县内冬小麦–夏玉米、棉花、玉米、红薯、谷子和花生的种植面积分别为148.23 km²、9.34 km²、18.25 km²、2.31 km²、0.77 km²和3.98 km²。

Abstract: Extraction of agricultural planting structure is a fundamental process. The traditional manual survey and statistical methods for obtaining crop information were subjective and prone to sig-nificant errors, which could not meet the management requirements of agriculture in the context of the information era. In recent years, the scarcity of water resources has constantly hindered the development of irrigation agriculture. Different crops have significant differences in the amount of irrigation required throughout their growth period. Therefore, a rapid and accurate understanding of the planting structure is of great benefit to water-saving agriculture. In this study, Guangping County was selected as the research area, and Sentinel-2 satellite remote sensing images were used as the basic data. With the standard of manually collected and visually interpreted sample points, the random forest algorithm was employed to classify the planting structure in Guangping County. The results showed that the overall accuracy of the classification results was 89.95%, with a Kappa coefficient of 0.88. Both the overall accuracy and Kappa coefficient were higher than 0.85, indicating that the classification results of various crops were accurate. According to statistics, the planting areas of winter wheat-summer maize, cotton, maize, sweet potato, millet and peanut in Guangping County were 148.23 km², 9.34 km², 18.25 km², 2.31 km², 0.77 km² and 3.98 km² respectively.