Experimental implementation of loop closure detection using data dimensionality reduction by spectral method

摘要

This paper presents experimental results about loop closure detection in mobile robots through spectral description of images set and data dimensionality reduction. Both, spectral description and representation in low dimension depend heavily on the concept of dominant eigenvector. Integration between Matlab and ROS interface was exploited to perform our experiments. Besides, two environments were used: real and computationally simulated. Results have shown that the method is capable of performing correct loop closure detection at a significantly lower computation cost, when compared with those obtained by OpenCV library for visual analysis.