Comparative Analysis of Stairways Detection Based on RGB and RGB-D Image

Stairways detection from RGB and RGB-D stair image is the challenging work in the computer vision research area. The detection system provides topnotch solution with greater portability in assisting visually impaired people and guiding the autonomous navigation system at the smart environments in the real world. In this paper, a framework is introduced to compare the stairways detection performance based on the RGB and RGB-D image. Here, the stairways candidate region is detected through the geometrical feature of a stair, i.e., stair steps are appeared in the concurrent sorted order. This feature is extracted from the concurrent parallel horizontal edges. The concurrent parallel horizontal edges are extracted from the canny edge image. For that, an edge linking and non-candidate edge elimination procedure is utilized in this work. Those are the key contributions of this paper. The stairway region of interest (ROI) is detected by aforementioned unique geometric feature of a stair and recognized the up, down, and negative stair by the support vector machine (SVM). For that, LBP and One-dimensional depth features are extracted from the RGB and RGB-D image respectively and sent to the SVM for classify. The stair images captured under different lighting conditions have been used to test the proposed framework to evaluate the resultant accuracy of the system.