The particularity and complexity of microwave scattering mechanism bring great challenges to target interpretation based on SAR image. In recent years, deep learning method has been applied to SAR interpretation with good results. However, because of the special characteristics of SAR image, the deep learning networks should be adapted so as to get better results. This paper makes an extension to the deep semantic segmentation network to handle the complex valued SAR images. In complex valued domain, the network effectively utilizes phase information of SAR data and provides an advantage for efficient image interpretation. An experiment of landcover classification in polarimetric SAR data is carried out, which verifies the effectiveness of the proposed network.
Paper
Full text
Equivalent Complex Valued Deep Semantic Segmentation Network For SAR Images
Semantic Scholar · Engineering · 2019
Abstract
The particularity and complexity of microwave scattering mechanism bring great challenges to target interpretation based on SAR image. In recent years, deep learning method has been applied to SAR interpretation with good results. However, because of the special characteristics of SAR image, the deep learning networks should be adapted so as to get better results. This paper makes an extension to the deep semantic segmentation network to handle the complex valued SAR images. In complex valued domain, the network effectively utilizes phase information of SAR data and provides an advantage for efficient image interpretation. An experiment of landcover classification in polarimetric SAR data is carried out, which verifies the effectiveness of the proposed network.