Depth-Fusion Based on Gaussian Mixture Model for RGB-D Visual SLAM

摘要：

　　Stochastic noises and disturbances in depth image about a scene observed by a RGB-D camera degrade the performance of a visual simultaneous localization and mapping(SLAM)system.For enhancing perception robustness,we develop a real-time SLAM system which uses a RGB-D camera as its sole sensor modality.The depth uncertainty for RGB-D features is described by Gaussian mixture model(GMM).A frame-constrained depth-fusion approach is then proposed to obtain accurate depth information of the current frame using the past frames in a local window.Experiments performed on public RGB-D TUM dataset have showed the proposed system outperforms the ORB-SLAM2.

关键词： Depth fusion Gaussian mixture model Visual SLAM

作者: Zhaotong Ding Ran Huang Biao Hu

作者单位: College of Information Science and Technology,Beijing University of Chemical Technology,Beijing 100029,China

会议类型: 国际会议

会议名称: 2019中国智能自动化大会(CIA,2019)

会议地点: 江苏镇江

会议语种:英文

页码: 234-242

在线出版日期: 2019-09-20（万方平台首次上网日期，不代表论文的发表时间）