Comparison of detectors and distance metrics for pose estimation

Abstract

In many artificial vision applications, it is necessary to know the pose (rotation and translation) of the camera with respect to some object in the real world. To know this pose, many algorithms are based on the detection and matching of common points of interest in two or more images. For that reason, it is necessary to have adequate algorithms for point detection and a robust metric for pose estimation. This paper presents a comparative analysis of three of the most popular algorithms for point detection and two popular metrics. In the detectors, the robustness to geometric distortions, robustness to noise and processing speed were compared. In the metrics robustness to noise and processing speed were compared.