Loop Closure Detection for Mobile Robot based on Multidimensional Image Feature Fusion

Loop closure detection is a crucial part of VSLAM. However, the traditional loop algorithms are difficult to adapt complex and changeable scenes. In this paper, we fuse Gist features, semantic features appearance image detect closures quickly accurately. Firstly, take advantage fast extraction speed feature by using it screen candidate frames. Then, current frame semantically segmented obtain mask blocks various types objects, nodes constructed calculate similarity between them. Next, images calculated according shape blocks. Finally, based on similarity, calculation model can be built as basis for detection. Experiments carried out both public self-filmed datasets. The results show that our proposed algorithm in scene accurately when illumination, viewpoint object change.