Appearance-Based Outdoor Localization Using Group LASSO Regression

This paper presents appearance-based localization for an omni-directional camera that builds on a combination of the group Least Absolute Shrinkage and Selection Operator (LASSO) and the extended Kalman filter (EKF). A histogram that represents the population of the Speeded-Up Robust Features (SURF points) is computed for each image, the features of which are selected via the group LASSO regression. The EKF takes the output of the LASSO regression-based first localization as observations for the final localization. The experimental results demonstrate the effectiveness of our approach.