A Real-Time Dynamic Object Segmentation Framework for SLAM System in Dynamic Scenes
Jianfang Chang, Na Dong, Donghui Li
Abstract
To accurately detect dynamic objects in dynamic scenes, a detection framework equipped with visual based measurement methods has been proposed in this paper. Firstly, to segment dynamic objects in real time, the real-time instance segmentation network YOLACT (You Only Look At CoefficienTs) has been introduced. Secondly, the geometric constraints have been utilized to further filter the missing dynamic feature points outside the segmentation mask. The dense optical flow method with adaptive threshold has been introduced to detect the missing dynamic objects driven by humans. Thirdly, background inpainting strategy has been proposed to restore the features occluded by dynamic objects. In order to verify the effectiveness of the dynamic object detection, the proposed method has been embedded in the visual Simultaneous Localization and Mapping (SLAM) system to improve its performance in dynamic environments. Experiments performed on the TUM and KITTI dataset have proved that the proposed detection method has excellent performance in dynamic scenes, which is of great significance to improve the robustness of the SLAM system.