Comparative Evaluation of Random Forest and Fern Classifiers for Real-Time Feature Matching

Abstract

Feature or keypoint matching is a critical task in many computer vision applications, such as optical 3D reconstruction or optical markerless tracking. These applications demand very accurate and fast matching techniques. We present an evaluation and comparison of two keypoint matching strategies based on supervised classification for markerless tracking of planar surfaces. We have applied these approaches on an augmented reality prototype for indoor and outdoor design review.