Learning Visual Landmarks for Pose Estimation

We present an approach to vision-based mobile robot localization, even without an a-priori pose estimate. This is accomplished by learning a set of visual features called image-domain landmarks. The landmark learning mechanism is designed to be applicable to a wide range of environments. Each landmark is detected as a focal extremum of a measure of uniqueness and represented by an appearance-based encoding. Localization is performed using a method that matches observed landmarks to learned prototypes and generates independent position estimates for each match. The independent estimates are then combined to obtain a final position estimate, with an associated uncertainty. Quantitative experimental evidence is presented that demonstrates that accurate pose estimates can be obtained, despite changes to the environment.

Authors

• Robert Sim
• Gregory Dudek

Keywords

• Sections
• Computational efficiency
• Prototypes
• Mobile robots
• Encoding
• Robot vision systems
• Cameras
• Robustness
• Layout
• State-space methods
• Pose Estimation
• Prototype
• Position Estimation
• Mobile Robot
• Independent Estimates
• Accuracy Of Pose Estimation
• Computational Cost
• Local Search
• Training Images
• Kalman Filter
• Indoor Environments
• Prediction Set
• Configuration Space
• Set Of Estimates
• Attention Model
• Median Estimates
• Image Position
• Correct Orientation
• Triangulation Method
• Camera Pose
• Robot Pose
• Landmark Detection
• Constrained Environments
• Edge Elements
• Edge Density
• Grid Position
• Pose Changes
• Regional Environment