AUTHOR=Peng Qiongyan , Shen Ruoque , Dong Jie , Han Wei , Huang Jianxi , Ye Tao , Zhao Wenzhi , Yuan Wenping
TITLE=A new method for classifying maize by combining the phenological information of multiple satellite-based spectral bands
JOURNAL=Frontiers in Environmental Science
VOLUME=10
YEAR=2023
URL=https://www.frontiersin.org/journals/environmental-science/articles/10.3389/fenvs.2022.1089007
DOI=10.3389/fenvs.2022.1089007
ISSN=2296-665X
ABSTRACT=Introduction: Using satellite data to identify the planting area of summer crops is difficult because of their similar phenological characteristics.
Methods: This study developed a new method for differentiating maize from other summer crops based on the revised time-weighted dynamic time warping (TWDTW) method, a phenology-based classification method, by combining the phenological information of multiple spectral bands and indexes instead of one single index. First, we compared the phenological characteristics of four main summer crops in Henan Province of China in terms of multiple spectral bands and indexes. The key phenological periods of each band and index were determined by comparing the identification accuracy based on the county-level statistical areas of maize. Second, we improved the TWDTW distance calculation for multiple bands and indexes by summing the rank maps of a single band or index. Third, we evaluated the performance of a multi-band and multi-period TWDTW method using Sentinel-2 time series of all spectral bands and some synthetic indexes for maize classification in Henan Province.
Results and Discussion: The results showed that the combination of red edge (740.2 nm) and short-wave infrared (2202.4 nm) outperformed all others and its overall accuracy of maize planting area was about 91.77% based on 2431 field samples. At the county level, the planting area of maize matched the statistical area closely. The results of this study demonstrate that the revised TWDTW makes effective use of crop phenological information and improves the extraction accuracy of summer crops’ planting areas over a large scale. Additionally, multiple band combinations are more effective for summer crops mapping than a single band or index input.