Binary-Patterns Based Floor Recognition Suitable for Urban Scenes

摘要

Nowadays urban structures (lines, planes, spheres, etc.) recognition is a useful task under computer vision systems since it provides rich scene information that can be exploited to understand the scene. In this context, one popular trend is for urban planar structures recognition (floor/ground recognition), because they have consistent appearance under urban scenarios and they can be used to improve several computer vision applications performance, for example, in autonomous vehicle navigation, robotic control, 3D modeling, etc. In the current literature, there are several approaches for floor recognition. However, most previous work has low robustness under image degradations (blur, lighting changes, noise, etc.). One alternative to address the image degradation problems is the use of binary features (for example LBP features). In this work, we propose a new binary-patterns based floor recognition suitable for urban scenes. For that, we propose two analyses, first we consider the floor connection to increase the recognition and second we segment the recognition in floor surface sets to remove the floor misrecognition. Finally, experimental results demonstrated that the proposed method delivers high stability under different scenes and it has more recognition than previous work under floor recognition domain.