We propose a minimalistic corridor representation consisting of the orientation line (center) and the wall-floor boundaries (lateral limit). The representation is extracted from low-resolution images using a novel combination of information theoretic measures and gradient cues. Our study investigates the impact of image resolution upon the accuracy of extracting such a geometry, showing that accurate centerline and wall-floor boundaries can be estimated even in texture-poor environments with images as small as 16 by 12. In a database of 7 unique corridor sequences for orientation measurements, less than 2% additional error was observed as the resolution of the image decreased by 99%. One of the advantages of working at such resolutions is that the algorithm operates at hundreds of frames per second, or equivalently requires only a small percentage of the CPU.
Orientation Line Estimation:
Median of bright pixels
Symmetry by mutual information
For orientation, we collected data for 4 different buildings, 8 unique corridors (1 training + 7 for testing). For every unique corridor, at equally spaced intervals along the corridor (4.5m), we rotated the robot from -20 degree to +20 degree and collected corresponding odometry (heading), laser readings and images. For wall-floor boundary and corridor reconstruction, we collected data for 11 distinct corridors in 6 different buildings. We drove the robot three times (middle, left, right separated by 0.5m) along each corridor and collected images along with their corresponding laser readings. The position of the orientation line with respect to the wall-floor boundaries gives the lateral position in the corridor. The distance between the two end-points in the wall-floor boundary yields the width of the corridor (in pixels). We use a homography transform obtained during a calibration process to recover corridor structure.
Orientation line estimation:
Three-line model estimation of the corridor geometry:
Corridor structure reconstruction from the wall-floor boundary, displayed as a top-down view:
Combined video of several different corridors
Yinxiao Li, Vidya Murali, and Stanley T. Birchfield, Extracting Minimalistic Corridor Geometry from Low-Resolution Images, Proceedings of the International Conference on Intelligent Robotics and Applications (ICIRA), Montreal, Quebec, Canada, October 2012.
This work was supported by NSF grant IIS-1017007.