The comparision of different visual features on RGB-D mapping

摘要

The availability of a map surrounding the robots' workspace is a significant requirement for the autonomous execution of the tasks, such as localization, navigation, and planning. RGB-D cameras can capture the RGB images along with per-pixel depth information from the environment. In this paper, we compare the influence of common visual features (SIFT, SURF, FAST, and GridFAST) with the latest ORB feature in RGB-D mapping system, which is built by combining the optimization algorithm with the visual features, in terms of the processing time, matching capability, and precision, and make exhaustive evaluations using the publicly available datasets. The results of experiments demonstrate that the performances of ORB feature, both in the processing time and the matching accuracy, are better than other features, and RGB-D mapping system based on ORB feature achieves the best real-time and relocalization, which is very important for robots to locate and navigate.