Line segment detection based on improved U-Net
Download as PDF
DOI: 10.23977/ESAC2020027
Author(s)
Yunpeng Zhi, Baolong Guo, Zhijie Chen and Cheng Li
Corresponding Author
Baolong Guo
ABSTRACT
Line segment detection measurement technology is an important and challenging task in computer vision, and it is also a hot research direction at home and abroad. This paper uses the improved U-Net network to extract line segments from simple textured images. The simple textured images no longer require high-level features, and pay more attention to the extraction of low-level features. The improved network enhances the low-level information in the image. Feature extraction has significantly improved the performance of straight line detection on images with simple textures.
KEYWORDS
LSD; U-Net network; Low-Level feature