Crop type mapping using time-series Sentinel-2 imagery and U-Net in early growth periods in the Hetao irrigation district in China

Authors: LI, GUANG - Northwest A&f University; CUI, JIAWEI - Northwest A&f University; HAN, WENTING - Northwest A&f University; Zhang, Huihui; HUANG, SHENGJIN - Northwest A&f University; CHEN, PAIPENG - Northwest A&f University; AO, JIANYI - Northwest A&f University

Submitted to: Computers and Electronics in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/30/2022
Publication Date: 11/8/2022
Citation: Li, G., Cui, J., Han, W., Zhang, H., Huang, S., Chen, P., Ao, J. 2022. Crop type mapping using time-series Sentinel-2 imagery and U-Net in early growth periods in the Hetao irrigation district in China. Computers and Electronics in Agriculture. 203. Article e107478. https://doi.org/10.1016/j.compag.2022.107478.
DOI: https://doi.org/10.1016/j.compag.2022.107478

Interpretive Summary: Satellite imagery can detect crop conditions and provide useful information for monitoring, mapping, and managing agricultural resources. However, it will be difficult to use these data until we know more about how data quality is affected by spatial or temporal resolution, image quality and cost, crop growth stage, and image classification techniques. In this study, we assessed new methods to extract information from Sentinel-2 satellite imagery to map wheat, corn, sunflower, or squash cropping areas as early as possible in the growing season in an irrigated region of Inner Mongolia, China. The proposed method could provide an efficient and reliable tool for crop type mapping and be used in other similar agricultural production areas.

Technical Abstract: The information on spatial distribution pattern and area of different crops is particularly useful for monitoring and managing the sustainability of agricultural resources. However, there are still challenges in timely mapping of crop types and planting areas to support production during the present growing season. Here, we proposed to use time-series Sentinel-2 imagery and a deep learning method for major crop mapping (wheat, maize, sunflower, and squash) in the agricultural production areas in the Hetao irrigation district, Inner Mongolia, China. A feature selection method that combines the global separability index and feature recursive elimination was proposed to optimize the band features of Sentinel-2 imagery. The random forest and U-Net deep learning algorithm were then used for image classification with all spectral bands and the selected band features, respectively. The selected bands and best method were used to determine the key time window for early identification of four major crops. The results showed that the proposed feature selection method effectively removed redundant features and improved the classification efficiency. The U-Net algorithm was more suitable for classification of major crops in the irrigation district with a higher accuracy whether using all band features or the selected features. The key periods for early recognition of wheat, maize, sunflower, and squash in the irrigation district were mid-May, late June, early July, and late June, respectively, and the F1-score, the mean intersection-over-union and overall accuracy were 69.02%, 68.37% and 81.26%, respectively. The proposed feature selection method, U-Net classification algorithm, and screening of best mapping window period in this study provide an efficient, accurate and timely crop mapping tool in the irrigation district. Future research can further verify the applicability of the early identification time window proposed in this study to support other agricultural management activities in the areas.