High-resolution remote sensing image road extraction method based on deformable attention mechanism

• Main Authors: DAI LING, ZHANG RONGTING, ZHANG GUANGYUN
• Format: Patent
• Language: Chinese; English
• Published: 24.05.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention provides a high-resolution remote sensing image road extraction method based on a deformable attention mechanism, and solves the problems of low complex road extraction accuracy and narrow application range in the field of high-resolution remote sensing. The method comprises the following implementation steps of: performing data enhancement on an existing high-resolution remote sensing image data set by using a sliding window, horizontal overturning and vertical overturning operations; forming a road enhancement module by utilizing one-dimensional convolution in four directions (horizontal, vertical, left diagonal and right diagonal); learning of deformable offset in the self-attention mechanism is constrained by using a 1 * 1 convolution module, and a deformable attention module is formed to automatically adapt to geometric transformation of a road; a road enhancement module and a deformable attention module are embedded between an encoder taking ResNet50 as a basic framework and a bilinear int