Visual simultaneous localization and mapping (SLAM) provides mapping and self-localization results for a robot in an unknown environment based on visual sensors, that have the advantages of small volume, low power con...Visual simultaneous localization and mapping (SLAM) provides mapping and self-localization results for a robot in an unknown environment based on visual sensors, that have the advantages of small volume, low power consumption, and richness of information acquisition. Visual SLAM is essential and plays a significant role in supporting automated and intelligent applications of robots. This paper presents the key techniques of visual SLAM, summarizes the current research status, and analyses the new trends of visual SLAM research and development. Finally, specific applications of visual SLAM in restricted environments, including deep space and indoor scenarios, are discussed.展开更多
Reconfigurable modular robots feature high mobility due to their unconstrained connection manners.Inspired by the snake multi-joint crawling principle,a chain-type reconfigurable modular robot(CRMR)is designed,which c...Reconfigurable modular robots feature high mobility due to their unconstrained connection manners.Inspired by the snake multi-joint crawling principle,a chain-type reconfigurable modular robot(CRMR)is designed,which could reassemble into various configurations through the compound joint motion.Moreover,an illumination adaptive modular robot identification(IAMRI)algorithm is proposed for CRMR.At first,an adaptive threshold is applied to detect oriented FAST features in the robot image.Then,the effective detection of features in non-uniform illumination areas is achieved through an optimized quadtree decomposition method.After matching features,an improved random sample consensus algorithm is employed to eliminate the mismatched features.Finally,the reconfigurable robot module is identified effectively through the perspective transformation.Compared with ORB,MA,Y-ORB,and S-ORB algorithms,the IAMRI algorithm has an improvement of over 11.6%in feature uniformity,and 13.7%in the comprehensive indicator,respectively.The IAMRI algorithm limits the relative error within 2.5 pixels,efficiently completing the CRMR identification under complex environmental changes.展开更多
由于视觉SLAM(Simultaneous Localization and Mapping)算法研究多建立于静态环境中,使得在动态环境下的应用造成较大定位偏移,极大降低了系统的稳定性。针对该问题,该文在原有视觉SLAM算法的基础上结合深度学习方法,对环境可能存在的...由于视觉SLAM(Simultaneous Localization and Mapping)算法研究多建立于静态环境中,使得在动态环境下的应用造成较大定位偏移,极大降低了系统的稳定性。针对该问题,该文在原有视觉SLAM算法的基础上结合深度学习方法,对环境可能存在的动态目标进行特征点剔除,从而提升系统在动态环境下的鲁棒性。采用的视觉SLAM系统为ORB-SLAM3,深度学习方法为YOLOv5的实例分割算法,采用对目标模型mask轮廓内特征点的检测算法及多视角几何方法进行特征点剔除。首先利用并行通信,将SLAM系统获取到的帧数据传入YOLOv5系统中进行可能为动态目标的分割,然后将其分割结果传回SLAM系统进行跟踪建图。同时改进词袋加载模型,提升加载速度,最终构建动态环境的稠密地图,具备可靠的实时性。通过在TUM数据集上的实验评估,该方法对比原SLAM框架及现阶段经典动态环境研究均有提升,其在保证平均帧率不降低的前提下精度较ORB-SLAM3的RMSE平均提升近89%。实验结果表明,对动态环境下的视觉SLAM算法有效改进,极大提升了系统的鲁棒性及稳定性。展开更多
基金The National Key Research and Development Program of China (2016YFB0502102)The National Natural Science Foundation of China (41471388).
文摘Visual simultaneous localization and mapping (SLAM) provides mapping and self-localization results for a robot in an unknown environment based on visual sensors, that have the advantages of small volume, low power consumption, and richness of information acquisition. Visual SLAM is essential and plays a significant role in supporting automated and intelligent applications of robots. This paper presents the key techniques of visual SLAM, summarizes the current research status, and analyses the new trends of visual SLAM research and development. Finally, specific applications of visual SLAM in restricted environments, including deep space and indoor scenarios, are discussed.
基金supported by the National Key R&D Program of China(Grant No.2018YFB1304600)the National Natural Science Foundation of China(Grant No.62003337)+1 种基金the Open Fund for State Key Laboratory of Robotics(Grant No.2023O03)the Liaoning Province Joint Open Fund for Key Scientific and Technological Innovation Bases(Grant No.2021-KF-12-05).
文摘Reconfigurable modular robots feature high mobility due to their unconstrained connection manners.Inspired by the snake multi-joint crawling principle,a chain-type reconfigurable modular robot(CRMR)is designed,which could reassemble into various configurations through the compound joint motion.Moreover,an illumination adaptive modular robot identification(IAMRI)algorithm is proposed for CRMR.At first,an adaptive threshold is applied to detect oriented FAST features in the robot image.Then,the effective detection of features in non-uniform illumination areas is achieved through an optimized quadtree decomposition method.After matching features,an improved random sample consensus algorithm is employed to eliminate the mismatched features.Finally,the reconfigurable robot module is identified effectively through the perspective transformation.Compared with ORB,MA,Y-ORB,and S-ORB algorithms,the IAMRI algorithm has an improvement of over 11.6%in feature uniformity,and 13.7%in the comprehensive indicator,respectively.The IAMRI algorithm limits the relative error within 2.5 pixels,efficiently completing the CRMR identification under complex environmental changes.
文摘由于视觉SLAM(Simultaneous Localization and Mapping)算法研究多建立于静态环境中,使得在动态环境下的应用造成较大定位偏移,极大降低了系统的稳定性。针对该问题,该文在原有视觉SLAM算法的基础上结合深度学习方法,对环境可能存在的动态目标进行特征点剔除,从而提升系统在动态环境下的鲁棒性。采用的视觉SLAM系统为ORB-SLAM3,深度学习方法为YOLOv5的实例分割算法,采用对目标模型mask轮廓内特征点的检测算法及多视角几何方法进行特征点剔除。首先利用并行通信,将SLAM系统获取到的帧数据传入YOLOv5系统中进行可能为动态目标的分割,然后将其分割结果传回SLAM系统进行跟踪建图。同时改进词袋加载模型,提升加载速度,最终构建动态环境的稠密地图,具备可靠的实时性。通过在TUM数据集上的实验评估,该方法对比原SLAM框架及现阶段经典动态环境研究均有提升,其在保证平均帧率不降低的前提下精度较ORB-SLAM3的RMSE平均提升近89%。实验结果表明,对动态环境下的视觉SLAM算法有效改进,极大提升了系统的鲁棒性及稳定性。
