Abstract:
This dataset consists of 190,832 manually-digitized cropland field boundaries, with associated attributes, within Brazil, Ukraine, United States of America, Canada, and Russia. Specifically, 22 regions of various sizes (74km2 - 38,000km2) spanning 5 countries were digitized over a range of predominant crop types over different time periods. These field boundaries were drawn over 20 m Sentinel-2 imagery. This field boundary dataset is a byproduct of a larger effort to map cropland burned area (Global Cropland Area Burned: GloCAB product [1]), however, it has several benefits beyond its original intent, including as a training dataset for machine-learning field size analyses, or a dataset to derive cropland field characteristics across different predominant crop types and geographies.
摘要:
该数据集由190,832个手动数字化农田边界组成,具有关联的属性,在巴西,乌克兰,美利坚合众国,加拿大,和俄罗斯。具体来说,跨越5个国家的22个不同大小的地区(74km2-38,000km2)在不同时间段内对一系列主要作物类型进行了数字化。这些场边界是在20mSentinel-2图像上绘制的。这个田地边界数据集是绘制耕地烧毁面积的更大努力的副产品(全球耕地烧毁面积:GloCAB产品[1]),然而,它有几个超出其初衷的好处,包括作为机器学习领域大小分析的训练数据集,或数据集,以得出不同主要作物类型和地理位置的农田特征。