Although shortwave infrared (SWIR) imaging provides superior tissue penetration and reduced autofluorescence for preclinical applications, quantitative fluorescence analysis is hindered by the limited dynamic range (DR) of InGaAs cameras, forcing a focus on either bright or dim anatomical features. We develop a high dynamic range (HDR) imaging method specifically adapted for the high-noise characteristics of InGaAs detectors to enable quantitative fluorescence imaging across wide intensity ranges. We demonstrate that one-time camera calibration based on a series of images encompassing the range of radiance intensities enables all subsequent image processing. We modified classical HDR algorithms with exposure-time-dependent dark current subtraction, preprocessing to exclude saturated and noisy pixels before camera response function recovery, and dynamic weighting range adjustment to account for shrinking intensity ranges at longer exposures. High dynamic range image processing effects on preclinical imaging outcomes were analyzed using indocyanine green and SWIR-emitting PbS/CdS quantum dots in mouse models. High dynamic range imaging achieved a 22-dB improvement in DR over single exposures, enabling simultaneous quantification across more than three orders of magnitude of fluorophore concentration. In vivo studies showed improvements in contrast-to-noise ratios across all anatomical features, with improvements in vascular contrast while maintaining quantitative accuracy. After one-time camera calibrations, this approach enables rapid processing of subsequent datasets. This software-based HDR SWIR imaging approach eliminates exposure parameter optimization and enables comprehensive biodistribution analysis across all anatomical structures from a single acquisition sequence, significantly streamlining preclinical imaging workflows while preserving quantitative accuracy.

Influence of the camera calibration process on the accuracy of vision-based 3D reconstruction for human gait analysis.

Accurate three-dimensional ear-position tracking is a prerequisite for ear-side active noise control in vehicle cabins, because natural head motion can directly shift the control target region and degrade the spatial consistency between the actual ears and the modeled control points. To address this problem, this study proposes a monocular-vision-based method for binaural three-dimensional localization and dynamic tracking in cabin scenes. A unified mapping among the pixel coordinate system, camera coordinate system, and cabin coordinate system is first established through camera calibration and geometric modeling. Facial landmark detection is then used to infer the two-dimensional locations of the left and right ears, after which cabin-feature-constrained monocular depth estimation is introduced to recover ear-region depth in a spatially aligned manner. The binaural three-dimensional coordinates are further refined through head-pose compensation, temporal filtering, and short-term prediction, so that the final output can be directly used as a control-oriented ear-state sequence. Multi-condition experiments under different illumination, occlusion, and head-pose variations show that the proposed method maintains good localization accuracy and trajectory continuity in all six tested cases. The root-mean-square error of binaural three-dimensional localization ranges from 18.40 to 25.57 mm, while the mean interaural distance error remains within 0.37 to 3.33 mm. Even under adverse conditions such as weak illumination and partial occlusion, the processed ear-state output remains continuously available to the downstream active noise control interface. These results indicate that the proposed method provides a low-cost and practically deployable solution for monocular binaural ear tracking in vehicle cabins and can serve as an effective perception front-end for ear-side active noise control.

Accurate measurement of the spatial position of targets in a fixed camera is critical in remote sensing applications. Visual spatial positioning methods that rely solely on images are susceptible to adverse factors such as inaccurate camera calibration, imprecise image target detection, and incorrect feature point selection. Complementary to images, the ubiquitous Global Navigation Satellite System (GNSS) data can provide spatial positions of targets, but most of them are low-cost GNSSs with significant positioning noise. In order to fuse these two valuable but flawed positioning measurements to improve the accuracy and stability of spatial positioning, we propose a deep learning multi-modal spatial positioning method by fusing sequential uncalibrated video images and low-cost GNSSs. Firstly, a self-supervised cascade denoising auto-encoder (SCDAE) architecture is built to endow the auto-encoder with robustness to noise in the raw inputs. Then, based on the SCDAE and Bayesian optimal estimation, a Bayesian self-supervised multi-modal fusion positioning method SCDAE-MFP is presented to achieve accurate and stable spatial positioning by self-supervised manifold learning. Specifically, to provide visual self-supervision to the SCDAE-MFP, a visual position denoising auto-encoder module based on dual unsupervised learning is proposed. Extensive experimental results on public datasets showed that SCDAE-MFP outperformed five other classical and state-of-the-art baseline methods by an average of 56.79% in reducing positioning errors.

Handball performance analysis is still often conducted through the manual review of match videos, while automation on broadcast footage remains challenging due to camera motion, strong perspective effects, and frequent occlusions during dense interactions. This study presents a practical and reproducible monocular pipeline for extracting handball analytics from a single broadcast viewpoint. Players are detected per frame, tracked over time, and projected onto a standardized handball court via homography-based camera calibration. The resulting court-referenced trajectories in metric units enable motion indicators such as distance covered and speed, along with coaching-oriented visual summaries, including trajectory overlays and heatmaps. In addition, clip-level action recognition is performed using interpretable kinematic and scene-derived features and lightweight classifiers, with a comparative evaluation across multiple classical models. The modular design keeps the intermediate steps explicit, supports reproducibility, and facilitates interpretation of both intermediate outputs and final analytics. Experiments on the UNIRI handball dataset demonstrate that meaningful performance analytics and action understanding can be obtained from single-camera broadcast video using transparent intermediate representations. This work highlights the practical potential of interpretable trajectory-based modeling for under-instrumented sports and provides a reproducible baseline for future extensions incorporating richer contextual cues.

High-resolution photogrammetry has become an influential non-invasive procedure of the non-erroneous 3D recreation of the conventional sculptural pieces, which has numerous benefits compared to the traditional casting and hand modeling procedures. The paper is a proposal of a high-fidelity photogrammetry system to document complex geometric characteristics of textures on the surfaces of iconic sculptures. The design incorporates the high-resolution imaging sensors, controlled illumination scenes, and optimized multi-angle image capture plans in order to provide maximum coverage and minimum reconstruction error. A systematic pipeline is adopted which includes camera calibration, feature detection by use of algorithms like SIFT, SURF and ORB and then an efficient feature matching, sparse reconstruction and dense point cloud construction. Different sculptural artifacts with different materials, size and different levels of texture are tested experimentally under different environmental conditions. The findings reveal that they are better, higher in accuracy, more intense in preserving the surface details and complete improvements in reconstruction than the traditional and baseline digital approaches. The suggested methodology ensures that human intervention is minimized but retains the cultural authenticity hence is very applicable in the digital archiving, restoration, online exhibition, and preservation of heritage. The paper will promote scalable, more accurate, and efficient 3D documentation methodologies in the cultural heritage preservation field.

Accurate and stable extrinsic calibration is the foundation of high-quality fusion sensing and positioning of camera and Light Detection and Ranging (LiDAR). However, traditional targetless calibration methods suffer from limitations such as poor scene adaptability and unstable convergence, which significantly restrict calibration accuracy and robustness in complex environments. Aiming at solving those problems, we propose an online cascade-optimization-based extrinsic calibration method of combining motion trajectory alignment and edge feature alignment. In the initial calibration stage, a hand-eye calibration algorithm is designed by minimizing the residual discrepancies between camera odometry and LiDAR odometry sequences. It establishes a robust initialization for subsequent optimization. Then, in order to extract robust edge line features from sparse point clouds, we employ depth difference and planar edges of point clouds in the optimization process. Subsequently, principal component analysis (PCA) is applied to compute the principal direction of the extracted line features, enabling a decoupled optimization scheme that accounts for directional observability. This approach effectively mitigates the adverse effects of uneven environmental feature distributions. Experimental validation on typical urban datasets demonstrates the method's generalizability and competitive accuracy: rotational parameter errors are constrained within 0.25°, and translational errors are maintained below 0.05 m. This affirms the method's suitability for high-accuracy engineering applications.

Improving 3D reconstruction quality for root phenotyping: assessing the impact of camera calibration and imaging parameters

Camera calibration in broadcast sports videos presents numerous challenges for accurate sports field registration due to multiple camera angles, varying camera parameters, and frequent occlusions of the field. Traditional search-based methods depend on initial camera pose estimates, which can struggle in non-standard positions and dynamic environments. In response, we propose an optimization-based calibration pipeline that leverages a 3D soccer field model and a predefined set of keypoints to overcome these limitations. Our method also introduces a novel refinement module that improves initial calibration by using detected field lines in a non-linear optimization process. This approach outperforms existing techniques in both multi-view and single-view 3D camera calibration tasks, while maintaining competitive performance in homography estimation. Extensive experimentation on real-world soccer datasets, including SoccerNet-Calibration, WorldCup 2014, and TS-WorldCup, highlights the robustness and accuracy of our method across diverse broadcast scenarios. Our approach offers significant improvements in camera calibration precision and reliability. Our project is available at https://github.com/mguti97/PnLCalib . • Most sports field registration methods focus on homography rather than full 3D camera calibration. • A robust keypoint grid based on field geometry can outperform existing methods. • Field’s landmark scarcity is mitigated by optimizing calibration with field lines.

Vibration measurement with neuromorphic vision sensors

Hyperspectral images 3D reconstruction based on structure-from-motion and multi-view stereo

Dataset of RGB-D images of object collections from multiple viewpoints with aligned high-resolution 3D models of objects.

A 3D point cloud dataset of Jining Qing Goats for segmentation analysis and body size measurement.

The external parameter calibration of LiDAR and camera is the core link of multimodal perception fusion technology, and its accuracy directly determines the effectiveness of cross modal information fusion. This article proposes a method for high-precision calibration of laser radar and camera extrinsic parameters using a checkerboard calibration board based on spatial registration theory. In response to the difficulty of feature extraction caused by the sparsity of LiDAR point clouds, this method enhances feature information by accumulating multi view point clouds and combines filtering algorithms to improve feature saliency; A standardized experimental platform was built based on ROS and Matlab, utilizing the reflection intensity distribution pattern of the checkerboard point cloud to achieve accurate extraction of its three-dimensional corner coordinates. At the same time, the corresponding two-dimensional checkerboard corner points are detected in the camera image, and the relative pose transformation matrix between the LiDAR and the camera is solved by establishing an accurate matching relationship between the two-dimensional and three-dimensional corner points. The experimental results show that this method has high-precision calibration results and good repeatability, which can provide a reliable extrinsic basis for multi-sensor fusion perception systems.

Given the insufficient weld image processing accuracy and poor tracking stability in complex industrial environments in laser vision automated welding, this paper constructs an improved U-Net architecture that uses multi-scale feature fusion to achieve weld image processing. Then, in the part of obtaining the coordinates of the weld target point, the visual tracking closed-loop is studied to introduce the hand-eye calibration link, and the position positioning and tracking of the weld seam image recognition are realized with the help of camera calibration. The results show that the improved U-Net has a peak signal-to-noise ratio of 42.5 decibels, a structural similarity index of 0.73 in high noise environments, and an edge preservation index of 0.82 in thick plate welding. The kernel correlation filter Kalman fusion algorithm achieves a tracking accuracy of 90.5% under compound motion mode and maintains a success rate of 86.7% under compound interference conditions. The positioning error of the laser vision system is only 0.69 mm when the calibration distance is 1,000 mm. Therefore, the proposed method is significantly superior to traditional methods in terms of denoising weld seam images, target tracking, and position localization.

Abstract Fringe projection profilometry has been widely applied in fields such as 3D measurement, industrial inspection, and automotive engineering due to its advantages of high precision, non-contact measurement, and simple structure. The calibration of fringe projection profilometry is a critical step for achieving high-precision measurements, as its accuracy directly affects the reliability of the measurement results. However, existing calibration methods usually require complex experimental setups or operational procedures, and they lack flexibility, making it challenging to adapt to diverse application scenarios. To address these issues, this paper proposes a novel and flexible calibration method for fringe projection profilometry with an auxiliary camera based on a phase-to-coordinate mapping model. Firstly, the phase-to-coordinate mapping model is derived by analysing geometric constraints in imaging and projecting process in fringe projection profilometry. With the aid of the auxiliary camera, the phase and 3D coordinates data are computed based on the stereo triangulation of the binocular vision built by the camera in fringe projection profilometry and the auxiliary camera after the camera calibration. After obtaining the phase and 3D coordinates, a random forest method is applied to remove the outliers of the phase-3D coordinate data, and the coefficients of the mapping model are computed by fitting the phase-to-coordinate mapping model. The proposed method reduces requirements of experimental conditions while improving the efficiency and adaptability of the calibration process. Experimental results demonstrate that the proposed method has advantages in calibration accuracy, stability, and operational convenience. It effectively meets the demands for efficiency and flexibility in practical industrial applications, providing a new solution for advancing fringe projection technology.

Accurate correction of radial lens distortion is a fundamental requirement in computer vision and photogrammetry, as geometric inaccuracies directly affect 3D reconstruction, mapping, and geospatial measurements, particularly in high-precision imaging systems. In this study, we propose a fully analytical, non-iterative method for truncated inverse modeling of radial lens distortion, applicable to general radial distortion polynomials that contain constant terms. Unlike classical truncated Lagrange series reversion, which relies on recursive expansion and combinatorial series construction, the proposed formulation determines inverse distortion coefficients directly through a system of constrained algebraic inverse polynomials. This enables deterministic computation of inverse parameters without iterative refinement, numerical root finding, or combinatorial complexity. The method was evaluated using ultra-wide-angle smartphone camera imagery exhibiting severe barrel distortion modeled by an eighth-degree depressed radial distortion polynomial. Its performance was compared with a commonly used iterative inverse modeling approach. The analytical formulation demonstrated improved numerical stability and substantially reduced reprojection errors when correcting highly nonlinear distortion profiles, achieving sub-pixel accuracy in image rectification. In contrast, the iterative approach exhibited instability and significantly larger reprojection errors under identical conditions. These results demonstrate that the proposed framework provides a general, robust, and repeatable solution for inverse radial distortion modeling, particularly for high-order polynomial models. The method offers clear practical advantages for camera calibration pipelines in photogrammetry, remote sensing, robotics, and other applications requiring high-fidelity imaging.

Single-view camera sensors suffer from inherent depth blurring issues that hinder progress in 3D human pose estimation (HPE), sparking widespread research interest in multi-view camera sensor systems. However, existing methods typically rely on complex camera calibration processes and are sensitive to dynamic environments. To address these limitations, we propose ECTFormer, an innovative calibration-free multi-view 3D HPE framework that seamlessly combines Transformer-based spatiotemporal modeling with CNN-based local feature extraction. The primary contents of this paper are: Firstly, we introduce a hierarchical multi-view spatio-temporal feature extraction network. This network avoids interference between noise from different viewpoints through hierarchical learning and employs a transformer to capture spatio-temporal features within views for subsequent fusion. And then, we design a CNN-Transformer fusion module (CTFM) that efficiently aggregates multi-view features, enabling accurate 3D pose regression. Extensive experiments on a public 3D human pose benchmark demonstrate that our approach attains superior performance without relying on calibration. For real-world environments equipped with dynamic camera sensors, ECTFormer can efficiently perform human pose estimation.

ABSTRACT To address the challenges of low mapping accuracy and target occlusion in multi-view roadside BEV perception, this paper proposes a weakly supervised bird's-eye view mapping framework. The method employs independent Transformer encoders to process images from each viewpoint and achieves accurate BEV alignment without per-view correspondence labels through cross-view consistency constraints, eliminating the need for dense BEV annotations or precise camera calibration. Experiments on the DAIR-V2X and a proprietary multi-view intersection dataset demonstrate that the framework outperforms homography-based mapping, BEVDet, and BEVFormer in terms of Point Mapping Squared Error (PMSE) and Overlap Rate (OR). Ablation studies further validate the effectiveness of the pairing loss and the independent mapping strategy. Although inference speed is slightly lower than homography-based methods, the proposed approach offers a practical balance between accuracy and efficiency for real-time roadside perception tasks.

IntroductionLiDAR–camera fusion systems are widely used in robotic localization and perception, where accurate extrinsic calibration is crucial for multi-sensor fusion. During long-term operation, extrinsic parameters can drift due to vibration and other disturbances, while target-based recalibration is inconvenient in the field and targetless approaches often suffer from highly non-convex objectives and limited robustness in challenging outdoor scenes.MethodsWe propose a targetless LiDAR–camera extrinsic calibration method by minimizing a semantic distribution consistency risk on SE(3). We align semantic probability distributions from the two sensing modalities in the image domain and freeze the pixel sampling measure at an anchor pose, so that pixel weighting no longer depends on the current extrinsic estimate and the objective landscape remains stable during optimization. On top of this anchor-fixed measure, we introduce a direction-aware weighting strategy that emphasizes pixels sensitive to yaw perturbations, improving the conditioning of rotation estimation. We further use a globally balanced Jensen–Shannon divergence to mitigate semantic class imbalance and enhance robustness.ResultsExperiments on the KITTI Odometry dataset show that the proposed method reliably converges from substantial initial perturbations and yields stable extrinsic estimates.DiscussionThe results indicate that the method is promising for maintaining long-term LiDAR–camera calibration in real-world robotic systems.