Abstract
Abstract. Most methods developed to map crop fields with high-quality are based on optical image time-series. However, often accuracy of these approaches is deteriorated due to clouds and cloud shadows, which can decrease the availably of optical data required to represent crop phenological stages. In this sense, the objective of this study was to implement and evaluate the conditional Generative Adversarial Network (cGAN) that has been indicated as a potential tool to address the cloud and cloud shadow removal; we also compared it with the Witthaker Smother (WS), which is a well-known data cleaning algorithm. The dataset used to train and assess the methods was the Luis Eduardo Magalhães benchmark for tropical agricultural remote sensing applications. We selected one MSI/Sentinel-2 and C-SAR/Sentinel-1 image pair taken in days as close as possible. A total of 5000 image pair patches were generated to train the cGAN model, which was used to derive synthetic optical pixels for a testing area. Visual analysis, spectral behaviour comparison, and classification were used to evaluate and compare the pixels generated with the cGAN and WS against the pixel values from the real image. The cGAN provided consistent pixel values for most crop types in comparison to the real pixel values and outperformed the WS significantly. The results indicated that the cGAN has potential to fill cloud and cloud shadow gaps in optical image time-series.
Highlights
Food demand is increasing rapidly and sustainable food production is currently a worldwide concern
It is possible to observe that the areas not affected by noises in the real images were more blurred in the conditional Generative Adversarial Network (cGAN) patches
It is possible to notice that the mean values for almost every cGAN band are closer to the real image than the Witthaker Smother (WS) image is, the clear exceptions are the RE2 and NIR bands for maize (Figure 3a) and the RE2 and RE3 bands for cerrado (Figure 3c) where the mean for the WS pixels are closer to the real image than the cGAN
Summary
Food demand is increasing rapidly and sustainable food production is currently a worldwide concern. Tillman et al (2011) predicted that by 2050 the world agricultural production should double to meet the food supply needs. In this context, stakeholders require regular and high-quality agricultural statistics to improve crop yield. The majority of existing approaches utilize optical image time-series to account for the crop phenological stages. Noises, clouds and cloud shadows in optical data can deteriorate the accuracy of land cover mapping. Lunetta et al (2006) pointed out the requirement of data cleaning pre-processing to remove the uncertainty and provide a better land cover mapping using time series. Shao et al (2016) outlined several smoothing algorithms that can be used to minimize noise in optical time series. The authors evaluated several algorithms and reported that Witthaker Smother (WS) provided the best distinction between classes of interest
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
