Discriminative and semantic feature selection for place recognition towards dynamic environments

Features play an important role in various visual tasks, especially place recognition applied to perceptually changing environments. We address challenges due dynamic and confusable patterns by proposing a discriminative semantic feature selection network named DSFeat this study. With supervision of both information attention mechanism, the pixel-wise stability features can be estimated, which indicates probability static region where are extracted. then select that insensitive interference distinguishable for correct matching. The designed model is evaluated SLAM system using several public datasets with varying appearances viewpoints. Experimental results demonstrate effectiveness proposed method. Note our method easily integrated into any feature-based system.