Efficient and fast stereo matching is a challenging task due to the presence of occlusion and low texture areas. In stereo matching, the correspondence between left and right images may be difficult owing to the lack of matching information. The cost metrics proposed before are not robust enough or are computationally expensive. In this work, we propose a novel generic tree structure, Pyramid-tree, which improves the single mode of traditional tree, and can achieve cross-regional connection between different regions with similar colors and similar depths. This unique structure can achieve cross-regional cost smoothing, significantly reducing the possibility of mismatch due to occlusion or lack of corresponding matching information, and has stronger robustness to occlusion and low texture regions. In addition, we also propose a new bearings-only cost metric, Log-angle, which is not affected by occlusion, low texture, illumination and other factors. Log-angle combines with traditional metrics can show better performance. We show that the Pyramid-tree structure and Log-angle are very important as it efficiently expands the state-of-the-art stereo matching methods and leads to significant improvements. Qualitative and quantitative experiments on Middlebury data sets verify the superior performance of the algorithm, and very effective compromise between the accuracy and computation load is achieved.
11 May 2023 in IEEE Transactions on Instrumentation and Measurement
Aiming at the challenging underwater navigation of autonomous underwater vehicle (AUV), this paper develops a novel filtering method with guided trajectory. The relatively low frequency of ultra-short baseline positioning system (USBL) measurement and the noise caused by complex scenes have a significant impact on the accuracy of AUV motion estimation. Although the existing adaptive extended Kalman filter (EKF) and adaptive unscented Kalman filter (UKF) which can roughly eliminate these outliers, the filtering results of these algorithms are not smooth and have the serious degree of the jitter, and then greatly affect the local consistency and quality of topographic map seamless stitching. We exploit the relatively smooth strap-down inertial navigation system (SINS) as the guided trajectory for motion estimation, and propose a trajectory smoothness preserving filter algorithm. This system has the advantage of maintaining local gradient consistency, which not only improves the navigation accuracy, but also ensures the smoothness of the estimation system. Simulation results show that our approach has high robustness and good smoothness. The effectiveness and practicability of the method are demonstrated by real experiments in multiple deep-sea areas with an average depth of more than 1000 meters in the southwest Indian Ocean.
This paper presents a new deep visual-inertial odometry and depth estimation framework for improving the accuracy of depth estimation and ego-motion from image sequences and inertial measurement unit (IMU) raw data. The proposed framework predicts ego-motion and depth with absolute scale in a self-supervised manner. We first capture dense features and solve the pose by deep visual odometry (DVO), and then combine the pose estimation pipeline with deep inertial odometry (DIO) by the extended Kalman filter (EKF) method to produce the sparse depth and pose with absolute scale. We then join deep visual-inertial odometry (DeepVIO) with depth estimation by using sparse depth and the pose from DeepVIO pipeline to align the scale of the depth prediction with the triangulated point cloud and reduce image reconstruction error. Specifically, we use the strengths of learning-based visual-inertial odometry (VIO) and depth estimation to build an end-to-end self-supervised learning architecture. We evaluated the new framework on the KITTI datasets and compared it to the previous techniques. We show that our approach improves results for ego-motion estimation and achieves comparable results for depth estimation, especially in the detail area.
5 September 2024 in IEEE Transactions on Vehicular Technology
Simultaneous localization and mapping (SLAM) based on LiDAR plays a pivotal role in many unmanned systems, but currently suffers from drift in trajectory estimation and lacks of robustness, resulting in inconsistent global maps. This paper proposes an accurate and robust LiDAR SLAM system to achieve low-drift ego-motion estimation and globally consistent mapping for unmanned ground vehicles (UGVs) in diverse environments. Firstly, a pyramid warm-up registration method is proposed to directly match the current scan with the map without feature extraction. More importantly, it utilizes the original geometric information to improve the registration accuracy and adopts a fast covariance matrix calculation method to greatly enhance the registration speed. Secondly, a submap generation method is proposed by formulating an anti-slip strategy and a point cloud similarity metric. It effectively prevents the loss of critical information while establishing strong constraints between keyframes and the map. Finally, a local-to-global optimization factor graph is constructed by establishing multi-level constraint relationships to optimize the overall system accuracy. The proposed method is compared with the current state-of-the-art LiDAR SLAM methods on several challenging datasets, including the KITTI, NeBula, and Newer College datasets. Experimental results show that our method has higher trajectory estimation accuracy and map consistency, and performs robustly in disparate environments
1 January 2026 in IEEE Transactions on Geoscience and Remote Sensing