Date of Award
8-2011
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Legacy Department
Electrical Engineering
Committee Chair/Advisor
BIRCHFIELD, STANLEY T
Committee Member
HOOVER , ADAM W
Committee Member
WALKER , IAN D
Committee Member
GREEN , KEITH E
Abstract
The goal of this research is to develop algorithms using low-resolution images to perceive and understand a typical indoor environment and thereby enable a mobile robot to autonomously navigate such an environment. We present techniques for three problems: autonomous exploration, corridor classification, and minimalistic geometric representation of an indoor environment for navigation.
First, we present a technique for mobile robot exploration in unknown indoor environments using only a single forward-facing camera. Rather than processing all the data, the method intermittently examines only small 32X24 downsampled grayscale images. We show that for the task of indoor exploration the visual information is highly redundant, allowing successful navigation even using only a small fraction (0.02%) of the available data. The method keeps the robot centered in the corridor by estimating two state parameters: the orientation within the corridor and the distance to the end of the corridor. The orientation is determined by combining the results of five complementary measures, while the estimated distance to the end combines the results of three complementary measures. These measures, which are predominantly information-theoretic, are analyzed independently, and the combined system is tested in several unknown corridor buildings exhibiting a wide variety of appearances, showing the sufficiency of low-resolution visual information for mobile robot exploration. Because the algorithm discards such a large percentage (99.98%) of the information both spatially and temporally, processing occurs at an average of 1000 frames per second, or equivalently takes a small fraction of the CPU.
Second, we present an algorithm using image entropy to detect and classify corridor junctions from low resolution images. Because entropy can be used to perceive depth, it can be used to detect an open corridor in a set of images recorded by turning a robot at a junction by 360 degrees. Our algorithm involves detecting peaks from continuously measured entropy values and determining the angular distance between the detected peaks to determine the type of junction that was recorded (either middle, L-junction, T-junction, dead-end, or cross junction). We show that the same algorithm can be used to detect open corridors from both monocular as well as omnidirectional images.
Third, 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 centerline and wall-floor boundaries can be estimated with reasonable accuracy even in texture-poor environments with low-resolution images. 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 99.9%.
Recommended Citation
Nariyambut murali, Vidya, "Low-Resolution Vision for Autonomous Mobile Robots" (2011). All Dissertations. 771.
https://open.clemson.edu/all_dissertations/771