Geometric Binary Descriptor Based Monocular SLAM

In this work, we develop an essential part of monocular simultaneous localization and mapping which is the matching process. However, the performance of any solution depends on the quality of the descriptor involved into the association approach. The most popular is binary descriptor due to their ability of matching with a high rate of matching in real time. In the present paper, we provide an ameliorated version of our previous work named local binary descriptor basing on 3D polynomial interpolation. The old version is reduced by the proposition of an algorithm of coefficients selection to deal with the real time constraint. This last is involved into a SLAM scheme named inverse depth parameterization for monocular SLAM. The result's quality is acceptable compared to the cross correlation in the old version of this SLAM.