2021年中国10米空间分辨率冬小麦识别数据集

及时准确的冬小麦空间分布信息对于粮食安全和作物生产管理至关重要。由于训练数据的获取成本高、效率低，大尺度、高质量、高空间分辨率的冬小麦分类产品依然匮乏。因此，我们提出了一种融合冬小麦物候、光谱和极化信息的训练数据自动化提取方法（Automated training data generation, ATDG），用于生成高质量的冬小麦训练样本，从而开展基于机器学习方法的冬小麦遥感制图。此外，基于生成的训练数据预训练分类模型，然后结合模型年际迁移方法（Model transfer, MT）实现了冬小麦生长季内制图。通过联合ATDG和MT，结合光学和雷达影像，我们生产了2020-2023年10米空间分辨率的中国冬小麦制图数据集（ChinaWheat10）。其中，2020、2021及2023年冬小麦分布图由ATDG方法生成，2022年分布图由MT方法生成。实地调查数据表明ChinaWheat10的总体精度在94%以上，在省市两级与统计数据的相关性（R2)分别在0.95和0.91以上，而且ChinaWheat10数据集中冬小麦田块细节明显。ATDG和MT方法在大尺度高分辨率高精度冬小麦自动化制图中具有巨大潜力。分类代码：1，冬小麦；0，非冬小麦。

Timely and accurate spatial distribution information of winter wheat is crucial for food security and crop production management. Due to the high acquisition cost and low efficiency of training data, winter wheat classification products with large-scale, high-quality and high spatial resolution are still scarce. Therefore, we proposed an automated training data generation (ATDG) method that fuses phenological, spectral and polarization information of winter wheat to generate high-quality winter wheat training samples, enabling winter wheat remote sensing mapping based on machine learning approaches. In addition, we pre-trained a classification model using the generated training data, and then combined with the inter-annual model transfer (MT) method to realize in-season mapping of winter wheat. By combining ATDG and MT, and integrating optical and radar imagery, we produced the ChinaWheat10 dataset, a 10-meter spatial resolution winter wheat mapping product of China spanning 2020 to 2023. Specifically, the winter wheat distribution maps for 2020, 2021 and 2023 were generated using the ATDG method, while the map for 2022 was generated via the MT method. Field survey data demonstrate that the overall accuracy of ChinaWheat10 exceeds 94%, and its correlation coefficients (R²) with statistical data at the provincial and municipal levels are above 0.95 and 0.91, respectively. Additionally, ChinaWheat10 exhibits distinct details of winter wheat fields. The ATDG and MT methods hold great potential for large-scale, high-resolution and high-precision automated mapping of winter wheat. Classification codes: 1 represents winter wheat; 0 represents non-winter wheat.

搜集汇总

背景与挑战

背景概述

该数据集是2021年中国10米空间分辨率的冬小麦识别数据集，通过融合物候、光谱和极化信息的自动化方法生成，结合光学与雷达影像实现高精度制图。数据总体精度超过94%，与统计数据高度相关，适用于大尺度冬小麦遥感监测和农业管理研究。