Abstract:
Aiming at the problem of false positive detection caused by current loop closure detection methods in visual SLAM (simultaneous localization and mapping), YOLOv3 (you look only once v3) object detection algorithm is adopted to obtain semantic information in scenes. Wrong and missing object detection are corrected by DBSCAN (density-based spatial clustering of application with noise) algorithm to create semantic nodes, which are then used to construct local semantic topology for a keyframe. After matching semantic nodes based on visual features and classification information of objects, transforming relationship can be computed for corresponding edges in different semantic topologies, which can obtain a similarity score. Judgement of loop closure is performed according to the changes of similarities between consequent keyframes. Experiments on benchmark datasets prove that object clustering effectively improves the accuracy of loop closure detection in indoor scenes. Compared with algorithms which are barely based on traditional visual features, the proposed algorithm can achieve loop closure detection with a higher accuracy.