ORB-SLAM Based Semi-Dense Mapping with Monocular Camera

In recent years, Simultaneous Localization And Mapping (SLAM) has been more and more popular especially in the field of unmanned vehicles and augmented reality (AR) applications. The existing feature based SLAM systems such as ORB-SLAM can only estimate sparse mapping and the direct methods such as LSD-SLAM have poor robustness on variety of light intensity. In this paper, we combine the advantages of direct method and indirect method, to accomplish semi-dense mapping based on ORB-SLAM. Our method has two key novelties: (1) when performing epipolar search, we compare the hamming distance of binary descriptor along the epipolar line, and thus the influence of illumination change can be reduced (2) we propose modified bilateral filter to smooth depth estimation. Experimental results show that a denser mapping can be obtained using our system.