NDVI Reconstruction by Synthesizing Optical and Synthetic Aperture Radar Images Based on CNN

As one of the most important vegetation index, NDVI can be used for extraction of phenological information and forest resource monitoring. However, optical data are often missing due to cloud cover, making it impossible to obtain time-series NDVI data. Synthetic aperture radar (SAR) has the capability of all-weather and all-day observation, SAR is also sensitive to vegetation growth. In this paper, a NDVI prediction model with multi-source images based on convolutional neural network is proposed, which achieves accurate reconstruction and prediction of NDVI in cloud-obscured regions by establishing relationship between SAR and optical images. The performance of the proposed SAR-optical synergistic NDVI prediction under different data inputs and model structures are are also evaluated. In addition, the NDVI construction results are also assessed for crop in different structure. The results show the potential of SAR in the reconstruction of time series NDVI.