Abstract
Ice-water detection is a major challenge for computer vision in the maritime environment. To address this challenge, we present a Convolutional Neural Network (CNN) model that fuses two imaging modalities: synthetic aperture radar (SAR) and advanced microwave scanning radiometer2 (AMSR2). The reasons for fusing these two imaging modalities are threefold. First, SAR provides high spatial resolution images, and AMSR2 provides images independent of wind conditions. In addition, the CNN fusion model can provide complementary information when images have different resolutions. Finally, the model generates a pixel-wise classification map for automatically generating sea ice charts, which reduces labour and time costs. We also investigate the effect of fusion on the segmentation performance by proposing uni-modal architecture, which is limited to the SAR modality. The results of this study show that the proposed model can accurately generate segmentation maps with more detailed sea ice textures and much sharper sea ice edges. The proposed fusion model achieves a pixel-wise accuracy of 94.60% and an F1-score of 94.99%.
Javad Sheikh
Project Researcher at
I’m an engineer and programmer passionate about solving problems with machine learning and deep learning.