In this paper a semi-direct visual odometry and mapping system is proposed with a RGB-D camera,which combines the merits of both feature based and direct based methods.The presented system directly estimates the camer...In this paper a semi-direct visual odometry and mapping system is proposed with a RGB-D camera,which combines the merits of both feature based and direct based methods.The presented system directly estimates the camera motion of two consecutive RGB-D frames by minimizing the photometric error.To permit outliers and noise,a robust sensor model built upon the t-distribution and an error function mixing depth and photometric errors are used to enhance the accuracy and robustness.Local graph optimization based on key frames is used to reduce the accumulative error and refine the local map.The loop closure detection method,which combines the appearance similarity method and spatial location constraints method,increases the speed of detection.Experimental results demonstrate that the proposed approach achieves higher accuracy on the motion estimation and environment reconstruction compared to the other state-of-the-art methods. Moreover,the proposed approach works in real-time on a laptop without a GPU,which makes it attractive for robots equipped with limited computational resources.展开更多
Aiming at the problem of system error and noise in simultaneous localization and mapping(SLAM) technology, we propose a calibration model based on Project Tango device and a loop closure detection algorithm based on v...Aiming at the problem of system error and noise in simultaneous localization and mapping(SLAM) technology, we propose a calibration model based on Project Tango device and a loop closure detection algorithm based on visual vocabulary with memory management. The graph optimization is also combined to achieve a running application. First, the color image and depth information of the environment are collected to establish the calibration model of system error and noise. Second, with constraint condition provided by loop closure detection algorithm, speed up robust feature is calculated and matched. Finally, the motion pose model is solved, and the optimal scene model is determined by graph optimization method. This method is compared with Open Constructor for reconstruction on several experimental scenarios. The results show the number of model's points and faces are larger than Open Constructor's, and the scanning time is less than Open Constructor's. The experimental results show the feasibility and efficiency of the proposed algorithm.展开更多
A fundamental task for mobile robots is simultaneous localization and mapping(SLAM).Moreover,long-term robustness is an important property for SLAM.When vehicles or robots steer fast or steer in certain scenarios,such...A fundamental task for mobile robots is simultaneous localization and mapping(SLAM).Moreover,long-term robustness is an important property for SLAM.When vehicles or robots steer fast or steer in certain scenarios,such as low-texture environments,long corridors,tunnels,or other duplicated structural environments,most SLAM systems might fail.In this paper,we propose a novel robust visual inertial light detection and ranging(Li Da R)navigation(VILN)SLAM system,including stereo visual-inertial Li Da R odometry and visual-Li Da R loop closure.The proposed VILN SLAM system can perform well with low drift after long-term experiments,even when the Li Da R or visual measurements are degraded occasionally in complex scenes.Extensive experimental results show that the robustness has been greatly improved in various scenarios compared to state-of-the-art SLAM systems.展开更多
针对传统ORB(Oriented FAST and Rotated BRIEF)算法提取图像特征时存在的特征点数量不足且分布不均匀问题,提出了一种基于四叉树的ORB特征阶梯分布算法。通过四叉树算法分割出特征点疏密不同的区域,对每个区域采用逐步降低阈值的方法,...针对传统ORB(Oriented FAST and Rotated BRIEF)算法提取图像特征时存在的特征点数量不足且分布不均匀问题,提出了一种基于四叉树的ORB特征阶梯分布算法。通过四叉树算法分割出特征点疏密不同的区域,对每个区域采用逐步降低阈值的方法,实现FAST(Features from Accelerated Segment Test)角点自适应提取;同时依据分割区域设置逐次递减的分割深度和特征点提取比例,以减少运算时间和特征冗余,使特征点分布更均匀。采用覆盖均匀度对特征点的均匀性进行量化。试验结果表明,该算法比传统ORB算法单幅图片的特征点提取数量平均多10.45%,覆盖均匀度平均低20%,运行时间比Mur-Artal算法平均减少20.54%,有效地提高了提取特征点的数量和均匀性,提升了运算效率。展开更多
基金Supported by the National Natural Science Foundation of China(61501034)
文摘In this paper a semi-direct visual odometry and mapping system is proposed with a RGB-D camera,which combines the merits of both feature based and direct based methods.The presented system directly estimates the camera motion of two consecutive RGB-D frames by minimizing the photometric error.To permit outliers and noise,a robust sensor model built upon the t-distribution and an error function mixing depth and photometric errors are used to enhance the accuracy and robustness.Local graph optimization based on key frames is used to reduce the accumulative error and refine the local map.The loop closure detection method,which combines the appearance similarity method and spatial location constraints method,increases the speed of detection.Experimental results demonstrate that the proposed approach achieves higher accuracy on the motion estimation and environment reconstruction compared to the other state-of-the-art methods. Moreover,the proposed approach works in real-time on a laptop without a GPU,which makes it attractive for robots equipped with limited computational resources.
基金Supported by the National Natural Science Foundation of China(61772379)
文摘Aiming at the problem of system error and noise in simultaneous localization and mapping(SLAM) technology, we propose a calibration model based on Project Tango device and a loop closure detection algorithm based on visual vocabulary with memory management. The graph optimization is also combined to achieve a running application. First, the color image and depth information of the environment are collected to establish the calibration model of system error and noise. Second, with constraint condition provided by loop closure detection algorithm, speed up robust feature is calculated and matched. Finally, the motion pose model is solved, and the optimal scene model is determined by graph optimization method. This method is compared with Open Constructor for reconstruction on several experimental scenarios. The results show the number of model's points and faces are larger than Open Constructor's, and the scanning time is less than Open Constructor's. The experimental results show the feasibility and efficiency of the proposed algorithm.
基金Project supported by the National Key R&D Program of China(No.2018YFB1305500)the National Natural Science Foundation of China(No.U1813219)。
文摘A fundamental task for mobile robots is simultaneous localization and mapping(SLAM).Moreover,long-term robustness is an important property for SLAM.When vehicles or robots steer fast or steer in certain scenarios,such as low-texture environments,long corridors,tunnels,or other duplicated structural environments,most SLAM systems might fail.In this paper,we propose a novel robust visual inertial light detection and ranging(Li Da R)navigation(VILN)SLAM system,including stereo visual-inertial Li Da R odometry and visual-Li Da R loop closure.The proposed VILN SLAM system can perform well with low drift after long-term experiments,even when the Li Da R or visual measurements are degraded occasionally in complex scenes.Extensive experimental results show that the robustness has been greatly improved in various scenarios compared to state-of-the-art SLAM systems.
文摘针对传统ORB(Oriented FAST and Rotated BRIEF)算法提取图像特征时存在的特征点数量不足且分布不均匀问题,提出了一种基于四叉树的ORB特征阶梯分布算法。通过四叉树算法分割出特征点疏密不同的区域,对每个区域采用逐步降低阈值的方法,实现FAST(Features from Accelerated Segment Test)角点自适应提取;同时依据分割区域设置逐次递减的分割深度和特征点提取比例,以减少运算时间和特征冗余,使特征点分布更均匀。采用覆盖均匀度对特征点的均匀性进行量化。试验结果表明,该算法比传统ORB算法单幅图片的特征点提取数量平均多10.45%,覆盖均匀度平均低20%,运行时间比Mur-Artal算法平均减少20.54%,有效地提高了提取特征点的数量和均匀性,提升了运算效率。
