Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Shaoquan Feng has a total of 42 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Robust regularization of UAV building contours using a corner-based density peak approach

    by Chong Wu, Tieding Lu, Shaoquan Feng, Ruisheng Wang, Wuyong Tao, Xijiang Chen, Bufan Zhao, Xianghong Hua


  • Enhancing GNSS RTK Positioning Performance in Challenging Urban Environments With Prior Road Surface Map

    by Xingxing Li, Jiacheng Qiu, Yuxuan Zhou, Shengyu Li, Chunxi Xia, Shaoquan Feng

    High-precision global navigation satellite system (GNSS) positioning is critical for autonomous vehicles and intelligent transportation systems. However, it faces significant challenges in urban canyons and signal-obstructed environments, where multipath effects and poor satellite geometry lead to accuracy degradation. In such cases, additional auxiliary information is required. Existing map-aided localization methods either focus on meter-level positioning through road network matching or turning to high-definition (HD) maps with higher maintenance costs. To address this, we propose a lightweight system that constructs road surface maps using only vehicle trajectories and integrates the map information with GNSS real-time kinematic (RTK) to enhance positioning accuracy. Our method leverages geometric constraints by iteratively optimizing the distance from the GNSS antenna to the road surface, applying elevation priors to mitigate vertical uncertainty and improve horizontal accuracy simultaneously, which also enables online calibration of antenna height differences. Experiments were conducted in both controlled test field environments and complex urban road scenario. The results demonstrate notable improvements in positioning accuracy in both horizontal and vertical dimensions. In complex and challenging scenarios, the use of prior road surface constraints led to average improvements of 59.43% in horizontal precision and 89.70% in vertical precision. The proposed road surface map, with its compact data and crowdsourcing compatibility, offers a cost-effective solution for high-precision urban navigation without reliance on expensive HD maps or high-end sensors.

  • Open Access

    SF-Loc: A Visual Mapping and Geo-Localization System based on Sparse Visual Structure Frames

    by Yuxuan Zhou, Xingxing Li, Shengyu Li, Chunxi Xia, Xuanbin Wang, Shaoquan Feng

    For high-level geo-spatial applications and intelligent robotics, accurate global pose information is of crucial importance. Map-aided localization is a universal approach to overcome the limitations of global navigation satellite system (GNSS) in challenging environments. However, current solutions face challenges in terms of mapping flexibility, storage burden and re-localization performance. In this work, we present SF-Loc, a lightweight visual mapping and map-aided localization system, whose core idea is the map representation based on sparse frames with dense but compact depth, termed as visual structure frames. In the mapping phase, multi-sensor dense bundle adjustment (MS-DBA) is applied to construct geo-referenced visual structure frames. The local co-visbility is checked to keep the map sparsity and achieve incremental mapping. In the localization phase, coarse-to-fine vision-based localization is performed, in which multi-frame information and the map distribution are fully integrated. To be specific, the concept of spatially smoothed similarity (SSS) is proposed to overcome the place ambiguity, and pairwise frame matching is applied for efficient and robust pose estimation. Experimental results on the cross-season dataset verify the effectiveness of the system. In complex urban road scenarios, the map size is down to 3 MB per kilometer and stable decimeter-level re-localization can be achieved. The code will be made open-source soon (https://github.com/GREAT-WHU/SF-Loc).

  • Open Access

    MASt3R-Fusion: Integrating Feed-Forward Visual Model with IMU, GNSS for High-Functionality SLAM

    by Yuxuan Zhou, Xingxing Li, Shengyu Li, Zhuohao Yan, Chunxi Xia, Shaoquan Feng

    Visual SLAM is a cornerstone technique in robotics, autonomous driving and extended reality (XR), yet classical systems often struggle with low-texture environments, scale ambiguity, and degraded performance under challenging visual conditions. Recent advancements in feed-forward neural network-based pointmap regression have demonstrated the potential to recover high-fidelity 3D scene geometry directly from images, leveraging learned spatial priors to overcome limitations of traditional multi-view geometry methods. However, the widely validated advantages of probabilistic multi-sensor information fusion are often discarded in these pipelines. In this work, we propose MASt3R-Fusion,a multi-sensor-assisted visual SLAM framework that tightly integrates feed-forward pointmap regression with complementary sensor information, including inertial measurements and GNSS data. The system introduces Sim(3)-based visualalignment constraints (in the Hessian form) into a universal metric-scale SE(3) factor graph for effective information fusion. A hierarchical factor graph design is developed, which allows both real-time sliding-window optimization and global optimization with aggressive loop closures, enabling real-time pose tracking, metric-scale structure perception and globally consistent mapping. We evaluate our approach on both public benchmarks and self-collected datasets, demonstrating substantial improvements in accuracy and robustness over existing visual-centered multi-sensor SLAM systems. The code will be released open-source to support reproducibility and further research (https://github.com/GREAT-WHU/MASt3R-Fusion).

  • SF-Loc: A Visual Mapping and Geo-Localization System Based on Sparse Visual Structure Frames

    by Yuxuan Zhou, Xingxing Li, Shengyu Li, Chunxi Xia, Xuanbin Wang, Shaoquan Feng

    For high-level geo-spatial applications and intelligent robotics, accurate global pose information is of crucial importance. Map-aided localization is a generally applicable approach to overcome the limitations of global navigation satellite system (GNSS) in challenging environments. However, current solutions face challenges in terms of mapping efficiency, storage burden, and re-localization performance. In this work, we present SF-Loc, a lightweight visual mapping and map-aided localization system, which is built upon the image-based map representation with dense but compact depth, termed as visual structure frames. In the mapping phase, multisensor dense bundle adjustment (MS-DBA) with global optimization is applied to construct geo-referenced visual structure frames. Besides, a co-visibility-based mechanism is developed to achieve incremental mapping and keep the map sparsity. In the localization phase, coarse-to-fine vision-based localization is performed, in which multiframe information and the map distribution are fully leveraged. To be specific, spatiotemporally associated similarity (SAS) is proposed to overcome the place ambiguity, and pairwise frame matching is applied for robust pose estimation. Experimental results on the real-world cross-temporal dataset verify the effectiveness of the system. In complex urban road scenarios, the map size is down to 3 MB per kilometer and stable decimeter-level relocalization can be achieved.


  • Open Access

    S3MOT: Monocular 3D Object Tracking with Selective State Space Model

    by Zhuohao Yan, Shaoquan Feng, Xingxing Li, Yuxuan Zhou, Chunxi Xia, Shengyu Li

    Accurate and reliable multi-object tracking (MOT) in 3D space is essential for advancing robotics and computer vision applications. However, it remains a significant challenge in monocular setups due to the difficulty of mining 3D spatiotemporal associations from 2D video streams. In this work, we present three innovative techniques to enhance the fusion and exploitation of heterogeneous cues for monocular 3D MOT: (1) we introduce the Hungarian State Space Model (HSSM), a novel data association mechanism that compresses contextual tracking cues across multiple paths, enabling efficient and comprehensive assignment decisions with linear complexity. HSSM features a global receptive field and dynamic weights, in contrast to traditional linear assignment algorithms that rely on hand-crafted association costs. (2) We propose Fully Convolutional One-stage Embedding (FCOE), which eliminates ROI pooling by directly using dense feature maps for contrastive learning, thus improving object re-identification accuracy under challenging conditions such as varying viewpoints and lighting. (3) We enhance 6-DoF pose estimation through VeloSSM, an encoder-decoder architecture that models temporal dependencies in velocity to capture motion dynamics, overcoming the limitations of frame-based 3D inference. Experiments on the KITTI public test benchmark demonstrate the effectiveness of our method, achieving a new state-of-the-art performance of 76.86~HOTA at 31~FPS. Our approach outperforms the previous best by significant margins of +2.63~HOTA and +3.62~AssA, showcasing its robustness and efficiency for monocular 3D MOT tasks. The code and models are available at https://github.com/bytepioneerX/s3mot.


  • LIO-LOT: Tightly-Coupled Multi-Object Tracking and LiDAR-Inertial Odometry

    by Xingxing Li, Zhuohao Yan, Shaoquan Feng, Chunxi Xia, Shengyu Li, Yuxuan Zhou

    Most existing LiDAR-Inertial odometry (LIO) systems heavily depend on the assumption of a static environment, which hinders their ability to effectively utilize dynamic objects in enhancing localization. Furthermore, accurate tracking of surrounding objects is crucial for emerging applications, such as autonomous driving and multi-robot collaboration. To this end, we present LIO-LOT, a tightly coupled LiDAR-inertial and multi-object tracking (MOT) system. In our system, each movable object is represented by a 3-D bounding box, derived from an object detector. Subsequently, these objects are associated with historical trajectories through a hybrid information matching strategy that integrates detections and geometric feature information. Considering tracking continuity and motion consistency, the associated objects are further categorized into high-confidence ones and low-confidence ones. Building upon this foundation, these two categories of objects are flexibly coupled with inertial measurement unit (IMU) pre-integration measurements and static scene structures to optimize the poses of both tracked objects and ego-vehicle under a factor graph optimization framework. For implementation, the aggregated information of high-and low-confidence objects are fully leveraged to form the object-related factors, which are hierarchically optimized with LIO, enhancing system stability in cases of poor object detections. The system’s performance is extensively evaluated using the KITTI and nuScenes datasets. Experimental results explicitly demonstrate that the proposed system outperforms other state-of-the-art methods in terms of ego-localization and multi-object tracking, particularly in highly dynamic scenarios with situations of occlusion and truncation.


  • Tightly Coupled Integration of LiDAR and Vision for 3D Multiobject Tracking

    by Shaoquan Feng, Xingxing Li, Zhuohao Yan, Chunxi Xia, Shengyu Li, Xuanbing Wang, Yuxuan Zhou

    Three-dimensional (3D) multiobject tracking (MOT) plays a pivotal role in emerging intelligent transportation system (ITS) applications, such as intelligent vehicles, wherein it provides accurate and reliable motion information of surrounding objects for safe navigation and motion planning. However, vision-based approaches are susceptible to occlusions and LiDAR-based methods struggle with object identification due to the lack of texture information, posing challenges in object tracking. To overcome the inherent limitations of these individual sensors, we propose a fusion-based 3D MOT system that tightly couples LiDAR and vision measurements. Specifically, we first design an aggregated association metric that fuses the detections and geometric features from both LiDAR and image domains, to effectively track objects. On this basis, all associated multimodal measurements are further integrated into a unified factor graph optimization framework to refine the tracked 2D and 3D trajectories. The extensive experiments conducted on public datasets demonstrate that our system significantly outperforms the state-of-the-art fusion-based MOT methods in terms of both object localization accuracy and MOT performance.