Mengyin Fu

Drones have become indispensable for reconstructing in-the-wild scenes due to their remarkable agility. Recent advances in radiance-field representations have achieved photorealistic rendering, opening new avenues for 3D reconstruction from drone imagery. However, limited view constraints hinder reliable geometric recovery, while dynamic distractors in the wild violate the multi-view consistency in radiance fields. In the wild, these two challenges are inextricably coupled. To address this coupled dilemma, we propose DroneSplat+, a novel framework that formulates semantics as a unifying hub to simultaneously resolve both issues. DroneSplat+ couples multiview stereo predictions with segmentation priors to impose semantic constraints that steer Gaussian optimization, enabling accurate geometry recovery under limited-view conditions. To suppress dynamics, DroneSplat+ introduces a dual-elimination strategy that combines adaptive local-global masking and labelbased localization to precisely identify and remove dynamic distractors from static scenes. We also provide a drone-captured 3D reconstruction benchmark dataset encompassing both dynamic and static scenes for comprehensive evaluation. Experimental results show that our method outperforms both 3DGS and NeRF baselines in reconstructing in-the-wild drone imagery. Project page: https://bityia.github.io/DroneSplat +/.

Capturing semantic and structural attributes of objects, together with their spatial relationships, is essential for achieving human-like scene understanding, thereby enhancing robotic tasks such as object association, loop closure detection, and map alignment in complex indoor environments. However, most existing object-based methods oversimplify geometric structures and spatial topology, resulting in coarse scene descriptions and inaccurate or even failed results in the above tasks. To address these limitations, this article presents a novel structural-semantic topological graph matching and map alignment method that fully utilizes scene structural, semantic, and topological cues to improve the performance of the above tasks. Specifically, we introduce a structural–semantic topological graph that jointly models object, plane, and line primitives to obtain a richer scene representation than object-only graphs. Then, we propose a novel instance association method that evaluates comprehensive matching similarity by jointly considering semantic, structural, and topological cues, while adopting a coarse-to-fine strategy to ensure accurate instance correspondences. Finally, we develop a robust pose estimation module that leverages a matches sampling strategy to effectively handle heterogeneous and potential noisy matches. Extensive experiments on the HM3D dataset and our real-world Hallway and Dyn-Room datasets demonstrate that the proposed method significantly outperforms the baselines in object association, loop closure detection, and map alignment.

Gravity-aided inertial navigation system (GAINS) is an effective passive navigation method, which represents significant progress in underwater navigation. The matching area selection algorithm is one of the key techniques. Distinctive from conventional methods, in this article, a nested method for matching area selection is proposed, which establishes a joint observation of gravity anomalies at the current positions and auxiliary positions to obtain more abundant observational information. Comprehensively considering factors such as sensors, gravity anomaly map, and trajectory, etc., this method measures the matching capacity from a system perspective rather than the 2-D image characteristics, and quantitatively estimates the matching biases. The possible maneuvering navigation and other changes of trajectory and attitude angles are focused on, and a preanalysis and real-time correction method is designed accordingly. Experimental results show that the proposed method can select matching area more accurately and efficiently, and makes it easier to set the matching area selection criteria than conventional algorithms.

Unmanned Aerial Vehicles (UAVs) autonomous exploration has shown great potential in emergency rescue, disaster relief and industrial inspection, etc. However, multi-scale environments, such as mines and buildings, pose great challenges to traditional methods sampling viewpoints at a fixed radius. Additionally, the disregard for informative gain and temporal constraints leads to back-and-forth maneuvers (BFMs). To address these problems, an efficient UAV autonomous exploration framework is developed using interactive viewpoints sampling. By introducing longitudinal and lateral sampling, the sampling method can enable the exploration framework to reduce back-and-forth maneuvers in multi-scale scenes. In addition, informative gain based guidance and temporal constraints are proposed. By balancing informative gains against path length and constraining adjacent moments of reverse maneuvers, the consistency of flight trajectories is improved. Both simulation and real world experiments were carried out to verify the effectiveness of proposed method. The results demonstrate that the proposed method outperforms other state-of-the-art approaches by 5.6% to 8.8% across all scenarios, including both general and multi-scale environments.

Object-level SLAM offers structured and semantically meaningful environment representations, making it more interpretable and suitable for high-level robotic tasks. However, most existing approaches rely on RGB-D sensors or monocular views, which suffer from narrow fields of view, occlusion sensitivity, and limited depth perception—especially in large-scale or outdoor environments. These limitations often restrict the system to observing only partial views of objects from limited perspectives, leading to inaccurate object modeling and unreliable data association. In this work, we propose MCOO-SLAM, a novel Multi-Camera Omnidirectional Object SLAM system that fully leverages surround-view camera configurations to achieve robust, consistent, and semantically enriched mapping in complex outdoor scenarios. Our approach integrates point features and object-level landmarks enhanced with open-vocabulary semantics. A semantic-geometric-temporal fusion strategy is introduced for robust object association across multiple views, leading to improved consistency and accurate object modeling, and an omnidirectional loop closure module is designed to enable viewpoint-invariant place recognition using scene-level descriptors. Furthermore, the constructed map is abstracted into a hierarchical 3D scene graph to support downstream reasoning tasks. Extensive experiments in real-world demonstrate that MCOO-SLAM achieves accurate localization and scalable object-level mapping with improved robustness to occlusion, pose variation, and environmental complexity.

Surround-view perception has garnered significant attention for its ability to enhance the perception capabilities of autonomous driving vehicles through the exchange of information with surrounding cameras. However, existing surround-view perception systems are limited by inefficiencies in unidirectional interaction pattern with human and distortions in overlapping regions exponentially propagating into non-overlapping areas. To address these challenges, this paper introduces ChatStitch, a surround-view human-machine co-perception system capable of unveiling obscured blind spot information through natural language commands integrated with external digital assets. To dismantle the unidirectional interaction bottleneck, ChatStitch implements a cognitively grounded closed-loop interaction multi-agent framework based on Large Language Models. To suppress distortion propagation across overlapping boundaries, ChatStitch proposes SV-UDIS, a surround-view unsupervised deep image stitching method under the non-global-overlapping condition. We conducted extensive experiments on the UDIS-D, MCOV-SLAM open datasets, and our real-world dataset. Specifically, our SV-UDIS method achieves state-of-the-art performance on the UDIS-D dataset for 3, 4, and 5 image stitching tasks, with PSNR improvements of 9%, 17%, and 21%, and SSIM improvements of 8%, 18%, and 26%, respectively. The code is available at https://github.com/lhlawrence/ChatStitch.

Robotic systems demand accurate and comprehensive 3D environment perception, requiring simultaneous capture of photo-realistic appearance (optical), precise layout shape (geometric), and open-vocabulary scene understanding (semantic). Existing methods typically achieve only partial fulfillment of these requirements while exhibiting optical blurring, geometric irregularities, and semantic ambiguities. To address these challenges, we propose OmniMap. Overall, OmniMap represents the first online mapping framework that simultaneously captures optical, geometric, and semantic scene attributes while maintaining real-time performance and model compactness. At the architectural level, OmniMap employs a tightly coupled 3DGS-Voxel hybrid representation that combines fine-grained modeling with structural stability. At the implementation level, OmniMap identifies key challenges across different modalities and introduces several innovations: adaptive camera modeling for motion blur and exposure compensation, hybrid incremental representation with normal constraints, and probabilistic fusion for robust instance-level understanding. Extensive experiments show OmniMap's superior performance in rendering fidelity, geometric accuracy, and zero-shot semantic segmentation compared to state-of-the-art methods across diverse scenes. The framework's versatility is further evidenced through a variety of downstream applications, including multi-domain scene Q&A, interactive editing, perception-guided manipulation, and map-assisted navigation.