Abstract:

For the problem of corridor recognition for autonomous mobile robot in the process of navigation in the indoor environment, an improved corridor recognition algorithm is proposed. The algorithm uses information of depth image provided by kinect sensor, and the depth difference between consecutive points and least squares are used to fit a straight line to detect the existence of corridor outlet. Experimental results show that the algorithm can efficiently recognize corridor types while having a robust, and is still able to maintain a high recognition rate of corridor types in a complex indoor environment.

Key words: corridor recognition; depth image; difference; least squares; Kinect