Introduction
及时、精确地获取冬小麦种植区域的重要性。遥感是一种主要手段。应用广泛的作物种植信息:光谱特征、时间特征、空间特征。大多机器学习方法只使用单幅影像,有局限。冬小麦与其他不过冬的作物相比,时序变化更加明显。简介 DTW。
研究区域的重要性。现有研究的不足。
Materials and Methods
1. Study Area
2. Data
2.1. Satellite Remote Sensing Data
Sentinel-2A/B MSI。Google Earth Engine。2020.8 ~ 2021.8
2.2. Field Survey Data
2.3. Administrative Division Data
2.4. Planting Data
3. Methods
3.1. Vegetation Index
3.2. Quality Mosaic Cloud Removal
3.3. S-G Filtering
3.4. DTW Algorithm
1、设 目标序列为,参考序列为
。其中,
分别是对应序列的第 n 和 m 个特征向量(eigenvector)。
2、两个点之间的距离为欧氏距离:
3、通过找出 T 和 R 的每个点之间的最小距离,找到一条从 T(N) 到 R(M) 的最小距离。该距离越小,T 和 R 越接近。建立一个阈值,距离小于阈值,就在识别为冬小麦。
3.5. Receiver Operating Characteristic (ROC) Curve
Process and Results
1. Data Processing and Results
1.1. Construction of the Time Series Vegetation Index
quality mosaic cloud removal、NDPI calculation、S-G filtering
1.2. determination of samples and reference curves
2. optimization of combination of characteristic phenological periods and their thresholds
3. Extraction of Winter Wheat in the Study Area
Discussion
1. Comparison of NDPI Curve Features of Time Series
2. Change in the Optimal Threshold
3. Change in Extraction Accuracy
4. Change in the Extracted Area from the Superposition of Cultivated Land Data