Simultaneous localization and mapping (SLAM) is one of the most essential technologies for mobile robots. Although great progress has been made in the field of SLAM in recent years, there are a number of challenges for SLAM in dynamic environments and high-level semantic scenes. In this paper, we propose a novel multimodal semantic SLAM system (MISD-SLAM), which removes the dynamic objects in the environments and reconstructs the static background with semantic information. MISD-SLAM builds three main processes: instance segmentation, dynamic pixels removal, and semantic 3D map construction. An instance segmentation network is used to provide semantic knowledge of surrounding environments in instance level. The ORB features located on the predefined dynamic objects are removed directly. In this way, MISD-SLAM effectively reduces the impact of dynamic objects to provide precise pose estimation. Then, combining multiview geometry constraint with K -means clustering algorithm, our system removes the undefined but moving pixels. Meanwhile, a 3D dense point cloud map with semantic information is reconstructed, which recovers the static background without the corruptions of dynamic objects. Finally, we evaluate MISD-SLAM by comparing to ORB-SLAM3 and the state-of-the-art dynamic SLAM systems in TUM RGB-D datasets and real-world dynamic indoor environments. The results indicate that our method significantly improves the localization accuracy and system robustness, especially in high-dynamic environments.

No abstract

Robot-CT is drawing great attention for its potential to support highly flexible scan trajectories and various scan modes with a single system. However, the quality of CT reconstruction largely depends on accurate knowledge of the scan geometry, i.e. position and orientation of the source and the detector. This imposes major challenges on current serial industrial robots. In order to realistically assess the impact of various robot characteristics on the Robot-CT reconstruction quality and to facilitate the development of a dedicated calibration method, this paper elaborates on a set of simulations based on the performance evaluation of an ABB industrial robot. Simulation results reveal that: a) synchronous error motions of the twin robots can partially compensate for the robot errors; b) offline calibration is a possible way to implement high accuracy Robot-CT due to the excellent repeatability of current industrial robots; c) both robot positioning and orientation errors need to be compensated for to get a high quality reconstruction; d) compensating for only the relative error cannot improve the final reconstruction quality: absolute positioning and orientation errors need to be measured and compensated for. External sensors or radiograph based absolute geometry calibration are needed to improve the reconstruction quality.

Robot CT systems bring great flexibility to X-ray CT based inspection scans. The system geometry needs to be well calibrated in order to avoid serious artifacts in the reconstructed volume. However, conventional calibration methods are normally designed for circular paths, which may be insufficient for a robot CT system. This paper proposes a phantom based geometric qualification method which supports nearly any view points. This is achieved by automatically mapping 3D-2D features based on collinear markers. The position and orientation of the X-ray focal spot and the detector are determined for each image of the reference phantom. This can be used for the CT reconstructions of later scans with the same trajectory. It is also possible to use the acquired geometries parameters to calibrate the underlying robots. The approach is validated on experimentally acquired data and the uncertainty is estimated by a Monte-Carlo based method.

No abstract