Detection And Localization Of Stair-cases Via Vision

The humanoid Robot have the capability which makes the autonomous mobile robot move in the human three-dimensional work space no longer constrained by two-dimensional plane. To the robot moving in the three-dimensional space, the stair may be the path to the destinations, or the obstacles need to avoid, or the marker used in navigation and localization. No matter what kind of application, the detection (or recognition) of stairs is the first and essential work. After recognizing the stairs the next work is the localization, evaluating the stairs location. Therefore detection and localization of the stair-cases target are the essential function of autonomous mobile robot system in any real and complex three-dimensional environmentIn this work the stair-cases detection and localization based on a single camera were researched, which contain the contents, edge detection and edge extraction, texture detection, stair-cases recognition algorithm and localization algorithm. Linear spatial filtering and Hough transform method were used to detect and extract the edge. The structural method of pattern recognition was used to identify the stair-cases. The image information and robot pose information and model parameters were used to build up the localization algorithm. The edge detection is unsuitable to the far distance stair-cases, so the Gabor filter was used as a supplement to texture detection algorithm.In this work the edge detection method was improved. For the problem of the adaptive threshold segmentation in edge detection algorithm, two solutions were put forward, the adaptive threshold segmentation based on the histogram shape and the adaptive threshold segmentation based on image enhancement. After experimental verification, the approach shows good adaptation for stairs identification in various circumstances.