A novel floor segmentation algorithm for mobile robot navigation

The task of detection of floor area for mobile robot navigation has received immense importance over the years. The main challenging problem as long as the path planning of robots is concerned, is the obstacle avoidance. Obstacle detection and avoidance in real time is a complex and computationally expensive process as a result of which the robotics researchers opted to segment out floors, which is comparatively easier process and at the same time very much feasible in real time applications. The proposed floor segmentation algorithm detects floor from a scene irrespective of the change of illumination as well as shadows. A conventional breadth first search based region growing technique has been used with histogram based features in YCbCr color space and floor junction masking to detect the floor pixels from the scene. The algorithm has been compared with various other existing techniques to showcase the improved results of the proposed one.