- 标题
- 摘要
- 关键词
- 实验方案
- 产品
-
[IEEE 2019 International Conference on Signal Processing and Communication (ICSC) - NOIDA, India (2019.3.7-2019.3.9)] 2019 International Conference on Signal Processing and Communication (ICSC) - Restoration of the Network for Next Generation (5G) Optical Communication Network
摘要： This paper describes the design, development and testing of an AR system that was developed for aerospace and ground vehicles to meet stringent accuracy and robustness requirements. The system uses an optical see-through HMD, and thus requires extremely low latency, high tracking accuracy and precision alignment and calibration of all subsystems in order to avoid mis-registration and “swim”. The paper focuses on the optical/inertial hybrid tracking system and describes novel solutions to the challenges with the optics, algorithms, synchronization, and alignment with the vehicle and HMD systems. Tracker accuracy is presented with simulation results to predict the registration accuracy. A car test is used to create a through-the-eyepiece video demonstrating well-registered augmentations of the road and nearby structures while driving. Finally, a detailed covariance analysis of AR registration error is derived.
关键词： see through HMD,hybrid tracking,augmented reality,image processing,calibration,registration,Inertial,sensor fusion
更新于2025-09-19 17:13:59
-
Review of mobile laser scanning target‐free registration methods for urban areas using improved error metrics
摘要： Registration is one of the most important tasks in mobile laser scanning (MLS) point cloud processing. This paper ?rstly reviews existing target-free matching techniques as well as methods to evaluate the quality of the registration. Next, a new error metric is introduced that takes into account the residuals of check planes as well as their orientation. Experiments using real datasets in combination with reference data were performed to evaluate the suitability of these metrics. The proposed error metric proved to be more suitable for evaluating the quality of point cloud registration than state-of-the-art equivalents. The results also indicate that least squares plane ?tting is the best technique for MLS point cloud registration.
关键词： point cloud matching,target-free registration,matching quality,error metric,mobile laser scanning
更新于2025-09-19 17:13:59
-
[IEEE 2019 IEEE International Instrumentation and Measurement Technology Conference (I2MTC) - Auckland, New Zealand (2019.5.20-2019.5.23)] 2019 IEEE International Instrumentation and Measurement Technology Conference (I2MTC) - A Method for Three-Dimensional Measurements Using Widely Angled Stereoscopic Cameras
摘要： Computer vision technologies have become popular tools for performing non-contact measurements. Stereoscopic systems have been used in several applications for length and geometry measurements. Three-dimensional (3D) reconstruction is an essential part of performing 3D measurements. A variety of methods have been developed for 3D reconstruction in stereoscopic systems. Block matching methods are considered as the most suitable option for 3D measurements, but they require the views to be similar for the cameras of a stereoscopic system. To satisfy this need, the cameras of a stereoscopic system should have small angles between their optical axes on the object’s surface. However, it is not always feasible nor desirable to arrange cameras in this way for some applications. We have proposed a new method to address this restriction. Our method uses an initial transform between the images from two cameras to make the views similar. Points on the transformed images are used as initial estimates of matched points in the two camera views. The points are then matched between the two images using an accurate subpixel image registration algorithm. The new method was tested using an object with known dimensions. The maximum measurement error achieved was 0.05 mm with a standard deviation of 0.09 mm for 10 measurements of a 12 mm length. The high accuracy of this method makes it a suitable option for applications that require reliable 3D measurements.
关键词： subpixel,3D reconstruction,block matching,stereoscopic measurements,wide base line cameras,image registration
更新于2025-09-16 10:30:52
-
Empirical stochastic model of detected target centroids: Influence on registration and calibration of terrestrial laser scanners
摘要： The target-based point cloud registration and calibration of terrestrial laser scanners (TLSs) are mathematically modeled and solved by the least-squares adjustment. However, usual stochastic models are simplified to a large amount: They generally employ a single point measurement uncertainty based on the manufacturers’ specifications. This definition does not hold true for the target-based calibration and registration due to the fact that the target centroid is derived from multiple measurements and its uncertainty depends on the detection procedure as well. In this study, we empirically investigate the precision of the target centroid detection and define an empirical stochastic model in the form of look-up tables. Furthermore, we compare the usual stochastic model with the empirical stochastic model on several point cloud registration and TLS calibration experiments. There, we prove that the values of usual stochastic models are underestimated and incorrect, which can lead to multiple adverse effects such as biased results of the estimation procedures, a false a posteriori variance component analysis, false statistical testing, and false network design conclusions. In the end, we prove that some of the adverse effects can be mitigated by employing the a priori knowledge about the target centroid uncertainty behavior.
关键词： stochastic model,target centroid detection,calibration,registration,terrestrial laser scanner
更新于2025-09-12 10:27:22
-
NRLI-UAV: Non-rigid registration of sequential raw laser scans and images for low-cost UAV LiDAR point cloud quality improvement
摘要： Accurate registration of light detection and ranging (LiDAR) point clouds and images is a prerequisite for integrating the spectral and geometrical information collected by low-cost unmanned aerial vehicle (UAV) systems. Most registration approaches take the directly georeferenced LiDAR point cloud as a rigid body, based on the assumption that the high-precision positioning and orientation system (POS) in the LiDAR system provides sufficient precision, and that the POS errors are negligible. However, due to the large errors of the low-precision POSs commonly used in the low-cost UAV LiDAR systems (ULSs), dramatic deformation may exist in the directly georeferenced ULS point cloud, resulting in non-rigid transformation between the images and the deformed ULS point cloud. As a result, registration may fail when using a rigid transformation between the images and the directly georeferenced LiDAR point clouds. To address this problem, we proposed NRLI-UAV, which is a non-rigid registration method for registration of sequential raw laser scans and images collected by low-cost UAV systems. NRLI-UAV is a two-step registration method that exploits trajectory correction and discrepancy minimization between the depths derived from structure from motion (SfM) and the raw laser scans to achieve LiDAR point cloud quality improvement. Firstly, the coarse registration procedure utilizes global navigation satellite system (GNSS) and inertial measurement unit (IMU)-aided SfM to obtain accurate image orientation and corrects the errors of the low-precision POS. Secondly, the fine registration procedure transforms the original 2D-3D registration to 3D-3D registration. This is performed by setting the oriented images as the reference, and iteratively minimizing the discrepancy between the depth maps derived from SfM and the raw laser scans, resulting in accurate registration between the images and the LiDAR point clouds. In addition, an improved LiDAR point cloud is generated in the mapping frame. Experiments were conducted with data collected by a low-cost UAV system in three challenging scenes to evaluate NRLI-UAV. The final registration errors of the images and the LiDAR point cloud are less than one pixel in image space and less than 0.13 m in object space. The LiDAR point cloud quality was also evaluated by plane fitting, and the results show that the LiDAR point cloud quality is improved by 8.8 times from 0.45 m (root-mean-square error [RMSE] of plane fitting) to 0.05 m (RMSE of plane fitting) using NRLI-UAV, demonstrating a high level of automation, robustness, and accuracy.
关键词： Low-cost,Light detection and ranging (LiDAR),Unmanned aerial vehicle (UAV),Image sequence,Non-rigid registration
更新于2025-09-11 14:15:04
-
[IEEE IGARSS 2018 - 2018 IEEE International Geoscience and Remote Sensing Symposium - Valencia, Spain (2018.7.22-2018.7.27)] IGARSS 2018 - 2018 IEEE International Geoscience and Remote Sensing Symposium - A Dense Vector Matching Approach for Band to Band Registration of Alsat-2 Images
摘要： The acquired images of the first Algerian high spatial resolution satellite Alsat-2 (A and B) images by the pushbroom scanner present a non-rigid misalignment among their different bands. The elimination of this registration error is a critical preprocessing step to assure the best use of this data for further applications. In this paper, a new approach for Alsat-2 band to band registration is proposed. This approach, called dense vector matching, leads to subpixel accuracies of the considered images. The proposed algorithm is based on a non-centered cross correlation technique. Indeed, it uses the following idea: when correlating each line-vector and each column-vector of the reference band with the neighboring vectors in the target band, subpixel misalignments are estimated by getting the maximum correlation after performing a cubic polynomial fit. Therefore, when applying the proposed algorithm to all bands, for each given pixel in the reference band, the new position of the corresponding pixel in the target bands is computed by subtracting the line and column registration errors from the reference position. Finally, the pixel-values of corrected target band are calculated by using the bilinear interpolation. Experiments are conducted to evaluate the proposed approach, and subpixel registrations are obtained on the corrected images, leading to an improvement of the multispectral images quality.
关键词： Alsat-2 images,dense vector matching,band to band registration
更新于2025-09-11 14:15:04
-
Extracting Accurate Building Information from Off-Nadir VHR Images
摘要： This research demonstrates the applicability of the improved algorithm for generating LoS-DSM elevation data through an elevation-based building detection in off-nadir VHR satellite imagery acquired over a dense urban area. The improved LoS-DSM algorithm was executed over a test dataset. The achieved image-elevation co-registration was very successful based on a visual assessment. Then, the generated and co-registered elevation data were applied in elevation-based building detection. The achieved building map was enhanced based on vegetation and occlusion masks as well as some morphological operations. The quality of the detection was evaluated based on manually generated reference data. The overall detection quality was found to be more than 90% with almost 95% of complete and correct detection. This level of performance in such a challenging dense urban area proves the high success of the disparity-based image-data co-registration as well as the applicability of the developed LoS-DSM elevations to detecting building objects even in off-nadir VHR satellite images acquired over dense urban areas.
关键词： off-nadir VHR images,urban areas,building detection,LoS-DSM,image-elevation co-registration
更新于2025-09-11 14:15:04
-
Accelerating multi-modal image registration using a supervoxel-based variational framework
摘要： For the successful completion of medical interventional procedures, several concepts, such as daily positioning compensation, dose accumulation or delineation propagation, rely on establishing a spatial coherence between planning images and images acquired at different time instants over the course of the therapy. To meet this need, image-based motion estimation and compensation relies on fast, automatic, accurate and precise registration algorithms. However, image registration quickly becomes a challenging and computationally intensive task, especially when multiple imaging modalities are involved. In the current study, a novel framework is introduced to reduce the computational overhead of variational registration methods. The proposed framework selects representative voxels of the registration process, based on a supervoxel algorithm. Costly calculations are hereby restrained to a subset of voxels, leading to a less expensive spatial regularized interpolation process. The novel framework is tested in conjunction with the recently proposed EVolution multi-modal registration method. This results in an algorithm requiring a low number of input parameters, is easily parallelizable and provides an elastic voxel-wise deformation with a subvoxel accuracy. The performance of the proposed accelerated registration method is evaluated on cross-contrast abdominal T1/T2 MR-scans undergoing a known deformation and annotated CT-images of the lung. We also analyze the ability of the method to capture slow physiological drifts during MR-guided high intensity focused ultrasound therapies and to perform multi-modal CT/MR registration in the abdomen. Results have shown that computation time can be reduced by 75% on the same hardware with no negative impact on the accuracy.
关键词： multi-modal registration,non-rigid registration,supervoxel,variational method
更新于2025-09-10 09:29:36
-
[Lecture Notes in Computer Science] Shape in Medical Imaging Volume 11167 (International Workshop, ShapeMI 2018, Held in Conjunction with MICCAI 2018, Granada, Spain, September 20, 2018, Proceedings) || Global Divergences Between Measures: From Hausdorff Distance to Optimal Transport
摘要： The data ?delity term is a key component of shape registration pipelines: computed at every step, its gradient is the vector ?eld that drives a deformed model towards its target. Unfortunately, most classical formulas are at most semi-local: their gradients saturate and stop being informative above some given distance, with appalling consequences on the robustness of shape analysis pipelines. In this paper, we build on recent theoretical advances on Sinkhorn entropies and divergences [6] to present a uni?ed view of three ?delities between measures that alleviate this problem: the Energy Distance from statistics; the (weighted) Hausdor? distance from computer graphics; the Wasserstein distance from Optimal Transport theory. The ε-Hausdor? and ε-Sinkhorn divergences are positive ?delities that interpolate between these three quantities, and we implement them through e?cient, freely available GPU routines. They should allow the shape analyst to handle large deformations without hassle.
关键词： Kernel,Optimal Transport,Energy Distance,Shape registration,Hausdor? distance,GPU
更新于2025-09-10 09:29:36
-
[IEEE 2018 IEEE 3rd International Conference on Image, Vision and Computing (ICIVC) - Chongqing (2018.6.27-2018.6.29)] 2018 IEEE 3rd International Conference on Image, Vision and Computing (ICIVC) - Registration of Point Clouds with Feature Extraction Based on Moving Least-Squares
摘要： In surface reconstruction, to solve the registration of point clouds of laser radar problem, a method based on moving least-squares was conducted to make feature extraction of target ball and then established linear equations to calculate the coordinate of the ball's center based on characteristic curves. Lastly, registration of point clouds was conducted based on four coordinates of the balls' centers. Experimental result shows that the method can improve the computing precision of the coordinate of the ball's center and the error of registration is in degree of millimeter based on moving least-squares. The accuracy is high and satisfies the engineering demand.
关键词： laser radar,reverse engineering,moving least-squares,registration of point clouds,feature extraction
更新于2025-09-10 09:29:36