camera

In multi-camera calibration for robotic vision systems, motion blur and rolling-shutter artifacts degrade image quality, leading to high re-capture rates and frequent manual intervention. To address this, we propose a voice-command-driven real-time precise image acquisition method. Our approach integrates a high-accuracy speech recognition model with millisecond-level timestamping and embedded microphone hardware to achieve sub-frame–level synchronization between voice triggering and image capture. Unlike conventional remote-control or post-hoc frame-filtering strategies, our method establishes an end-to-end temporally controllable calibration image acquisition pipeline. Experiments in complex multi-camera setups demonstrate substantial improvements: calibration success rate and robustness increase significantly, re-capture rate decreases by 62%, and overall calibration efficiency improves by 3.1×. This work establishes a reliable, user-friendly paradigm for on-site autonomous calibration of robotic vision systems.

Joint extrinsic calibration of event cameras, LiDAR, and RGB cameras remains challenging due to severe error accumulation and lack of effective multi-modal calibration targets. Method: This paper proposes an end-to-end joint calibration framework leveraging a custom-designed 3D multimodal calibration target—integrating planar geometry, ChArUco markers, and an active LED array—to enable synchronized, single-shot observation across all three modalities. The method jointly optimizes temporal, spatial, and geometric alignment by fusing geometric feature matching, ChArUco detection, spatiotemporal activation pattern recognition in event streams, and cross-sensor synchronization. Contribution/Results: Evaluated on a newly collected autonomous driving multimodal dataset, the framework achieves significantly higher extrinsic calibration accuracy than state-of-the-art pairwise methods, demonstrates strong robustness, and effectively resolves the long-standing instability issues in event camera calibration.

This paper addresses the relative pose estimation problem for three calibrated cameras given only four correspondences across all views. To overcome limitations of conventional methods—namely, their reliance on more correspondences or insufficient robustness—we propose a novel strategy that approximates a fifth correspondence using the centroid of the four observed points. We further introduce the first joint three-view pose estimation framework integrating a 4-point affine fundamental matrix solver, a standard 5-point relative pose solver, and a P3P solver. Geometric modeling enhances robustness against noise and outliers, while local optimization refines accuracy. Evaluated on real-world datasets, our method achieves state-of-the-art performance: the centroid-based strategy significantly outperforms pure affine approaches, striking a superior balance among accuracy, robustness, and computational efficiency, with straightforward implementation.

Accurate calibration of multi-camera systems remains challenging under realistic constraints such as the absence of calibration boards, inaccessible scenes, and unsynchronized video streams. To address this, we propose a two-stage calibration method that requires no physical calibration patterns. In the first stage, intrinsic parameters and effective field-of-view (EFOV) are estimated from a single static image per camera by manually annotating geometric primitives—namely, parallel, orthogonal, and equal-length line segments—exploiting natural scene geometry. In the second stage, extrinsic parameters are recovered via interactive EFOV plane projection and alignment with virtual calibration elements, enabling calibration without a physical planar target. The method significantly enhances flexibility and applicability, requiring only one static image per camera. Experimental results demonstrate superior accuracy and robustness compared to conventional calibration approaches. It has been successfully deployed in complex real-world scenarios for multi-camera collaborative tracking.

Existing video generation methods suffer from insufficient precision in camera motion control and neglect explicit modeling of subject motion dynamics, failing to meet professional-grade controllability requirements. To address this, we propose a high-precision, disentangled framework for joint camera and subject control. Our approach introduces 3D point trajectories in the camera coordinate system as control signals, explicitly modeling high-order motion dynamics—including acceleration and jerk—and incorporates an adjustable motion scaling operator. We adopt a lightweight, base-model-agnostic Adapter-based fine-tuning architecture. Extensive experiments demonstrate that our method significantly outperforms state-of-the-art approaches on both static and dynamic scenes. Quantitative evaluations show consistent improvements across key metrics (e.g., CAM-PSNR, Motion-FID), while qualitative results exhibit markedly more accurate camera choreography and natural, fine-grained controllability over subject motion.

This work addresses the challenges in extrinsic calibration between motion capture systems and external cameras—particularly fisheye cameras—in large-scale data collection, where errors arising from calibration board misalignment, ambiguous initialization, and temporal drift are difficult to detect promptly. To overcome these issues, the authors propose a robust joint calibration and independent validation framework that simultaneously estimates camera extrinsics and the transformation from the calibration board to motion capture markers. The method employs a staged nonlinear optimization strategy insensitive to initialization and introduces a fully independent validation pipeline—the “lollypop” module—that effectively handles non-uniform fisheye distortion. Experiments on the Meta Quest 3 demonstrate superior calibration accuracy over existing approaches, with the lollypop module reliably detecting calibration degradation during long-term operation. The system has been successfully deployed in real-world data acquisition pipelines.

This work addresses geometric distortions and visual artifacts commonly observed in monocular portrait video generation under controllable camera trajectories, which often stem from scale ambiguity or inaccuracies in 3D reconstruction. To overcome these limitations, the authors propose a face-aware, scale-aware camera representation that enables high-quality, temporally coherent, and controllable dynamic camera motion without relying on explicit 3D priors. The approach introduces a deterministic scale-aware camera conditioning mechanism and integrates synthetic camera motions with a multi-shot stitching strategy. It is trained within a large-scale video generation framework that jointly leverages multi-view studio data and in-the-wild monocular videos. Experiments demonstrate significant improvements over existing methods in camera controllability, visual fidelity, identity preservation, and motion authenticity on both the Ava-256 benchmark and diverse in-the-wild video datasets.

Existing video re-rendering methods often suffer from depth estimation artifacts in real-world dynamic scenes, struggling to simultaneously achieve appearance consistency and precise camera control. This work proposes a 4D point cloud–based re-rendering framework that anchors both the input video and the target camera trajectory into a unified 4D point cloud representation, explicitly preserving observed content while providing rich geometric priors. By integrating static pixel segmentation with multi-view dynamic reconstruction, the method significantly enhances robustness against real-world point cloud artifacts. Experiments demonstrate that the proposed framework consistently outperforms existing approaches across diverse videos and camera paths, achieving notable improvements in 4D temporal consistency, camera control accuracy, and visual fidelity, and successfully generalizes to complex real-world scenarios.

This work addresses the scalability limitation of conventional multi-projector calibration, which requires sequential projection of structured light patterns, resulting in a linear increase in calibration time with the number of projectors. To overcome this, the authors propose an embedded-camera calibration method that integrates an array of cameras into the calibration target to simultaneously capture structured light patterns from all projectors. By leveraging the direction of incident rays to disentangle overlapping patterns, the method establishes correspondences between the optical centers of the embedded cameras and projector pixels, enabling joint estimation of intrinsic and extrinsic parameters. This approach achieves, for the first time, synchronous multi-projector calibration using embedded cameras, reducing the projection–capture cycle from linear to nearly constant time. It maintains accuracy comparable to traditional methods while significantly improving calibration efficiency for large-scale, densely packed projector systems, with direct applicability to high-brightness blending, super-resolution, light-field displays, and shadow suppression.

This work addresses the challenge of extrinsic calibration for non-overlapping multi-camera systems by proposing a novel method that requires only pure rotational motion and a single static calibration target. By introducing an implicit turntable coordinate frame and formulating a 3D reprojection error on the SE(3) manifold, the approach integrates observations of the same calibration board captured by different cameras at distinct time instances into a unified global nonlinear optimization framework. The method eliminates the need for large calibration patterns or complex motion estimation, thereby avoiding scale ambiguity and drift issues. High accuracy, strong robustness, and ease of deployment are demonstrated on both controlled rigs and real-world vehicle platforms, marking the first successful realization of high-precision extrinsic calibration for non-overlapping multi-camera setups using only pure rotation and a single static target.