Year: 2025
Type: article
Abstract: Vector images have the advantages of small storage, easy preparation, and lossless transformation, and the image matching navigation method using road vector images as the base map has gained attention and research. We propose to use SAR images as real-time images and road vector images as reference images for matching, and prepare a SAR-VEC dataset. A Siamese U-Net dual-task supervised network is proposed. The U-Net with weight sharing is used to extract common road information between SAR and vector image images. The weighted sum of segmentation loss and matching loss is calculated as the network loss to jointly supervise network feature extraction. Experiments show that the method can obtain high matching correctness, with 80.2% correctness within 5 pixels of matching error and 91.0% correctness within 10 pixels of matching error. Compared with the existing methods, this method is able to identify the differences of similar roads and better eliminate the influence of imaging interference on the matching results in SAR images, obtaining more accurate matching results with better robustness.
Source: Remote Sensing
DOI: 10.3390/rs17203504
Authors: Hanyu Cai, Yong Xian, Shaopeng Li, Decao Ma
Language: English
Subfield: Ocean Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $3,258
OpenAlex ID: W4415402936