Road Detection via a Dual-Task Network Based on Cross-Layer Graph Fusion Modules

Road detection based on remote sensing images is of great significance to intelligent traffic management. The performances the mainstream road methods are mainly determined by their extracted features, whose richness and robustness can be enhanced fusing features different types cross-layer connections. However, in existing model frameworks often similar same layer single-task training, traditional fusion ways too simple obtain an efficient effect, so more complex besides concatenation addition deserve explored. Aiming at above defects, we propose a dual-task network (DTnet) for graph module (CGM): DTnet consists two parallel branches area edge detection, respectively, while enhancing feature diversity between through our designed bridge modules (FBM). CGM improves effect stream graph, four patterns evaluated. Experimental results three public datasets demonstrate that method effectively final result.