Robust registration of SAR and optical images based on deep learning and improved Harris algorithm

Autor: Wannan Zhang
Jazyk: angličtina
Rok vydání: 2022
Předmět:
Zdroj: Scientific Reports, Vol 12, Iss 1, Pp 1-8 (2022)
Druh dokumentu: article
ISSN: 2045-2322
DOI: 10.1038/s41598-022-09952-w
Popis: Abstract Traditional algorithms can achieve good results when registering homologous images, but it cannot reach satisfying results for registration between synthetic aperture radar (SAR) and optical images. The difficulty is that the image texture information and structures of different modalities is very different which leads to poor registration results. To solve this problem, we present a robust matching framework for registration between SAR and optical images. First, a novel deep learning network is utilized to generate high quality pseudo-optical images from SAR images. Next, feature points are detected and extracted using the multi-scale Harris algorithm. Then the feature points are constructed through the gradient position orientation histogram method. Finally, the actual position of the feature points will be reconstructed through a feedback mechanism for matching. Experimental results demonstrate its superior matching performance with respect to the state-of-the-art methods.
Databáze: Directory of Open Access Journals
Nepřihlášeným uživatelům se plný text nezobrazuje