In this paper, we reported the current result on the development of a cloud segmentation strategy for multispectral imageries (MSI) captured by LAPAN-A2 satellite. The segmentation was performed by involving 3200 sets of images using both deep-learning and classical-based approaches. For the deep-learning side, the U-Net was considered since it is one of the state-of-the-art image segmentation methods. On the other side, HSV (Hue, Saturation, Value) color space-based segmentation was chosen for the classical-based method. The performance of both approaches has been evaluated and compared in terms of their accuracy and speed. The comparison results provided in this paper could be used as a reference in choosing a proper strategy to extract cloud blobs existing on LAPAN-A2 MSI.
Cloud Segmentation Strategy for LAPAN-A2 Multispectral Imagery
2021-11-03
4323781 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch