Automatic LiDAR-Camera Calibration of Extrinsic Parameters Using a Spherical Target

Toth, Tekla; Pusztai, Zoltán; Hajder, Levente

doi:10.1109/icra40945.2020.9197316

Cited by 61 publications

(38 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Besides planar calibration targets, spherical targets are also appropriate for LiDARcamera calibration [31][32][33]. Compared with planar targets, the advantage of the spherical target is that the camera can automatically detect the outline without depending on the camera's angle of view and placement.…”

Section: Target-based Approachmentioning

confidence: 99%

See 1 more Smart Citation

CFNet: LiDAR-Camera Registration Using Calibration Flow Network

Wang

Dou

et al. 2021

Sensors

View full text Add to dashboard Cite

As an essential procedure of data fusion, LiDAR-camera calibration is critical for autonomous vehicles and robot navigation. Most calibration methods require laborious manual work, complicated environmental settings, and specific calibration targets. The targetless methods are based on some complex optimization workflow, which is time-consuming and requires prior information. Convolutional neural networks (CNNs) can regress the six degrees of freedom (6-DOF) extrinsic parameters from raw LiDAR and image data. However, these CNN-based methods just learn the representations of the projected LiDAR and image and ignore the correspondences at different locations. The performances of these CNN-based methods are unsatisfactory and worse than those of non-CNN methods. In this paper, we propose a novel CNN-based LiDAR-camera extrinsic calibration algorithm named CFNet. We first decided that a correlation layer should be used to provide matching capabilities explicitly. Then, we innovatively defined calibration flow to illustrate the deviation of the initial projection from the ground truth. Instead of directly predicting the extrinsic parameters, we utilize CFNet to predict the calibration flow. The efficient Perspective-n-Point (EPnP) algorithm within the RANdom SAmple Consensus (RANSAC) scheme is applied to estimate the extrinsic parameters with 2D–3D correspondences constructed by the calibration flow. Due to its consideration of the geometric information, our proposed method performed better than the state-of-the-art CNN-based methods on the KITTI datasets. Furthermore, we also tested the flexibility of our approach on the KITTI360 datasets.

show abstract

Section: Target-based Approachmentioning

confidence: 99%

“…Compared with planar targets, the advantage of the spherical target is that the camera can automatically detect the outline without depending on the camera's angle of view and placement. Besides, sampling points of spherical objects can be detected conveniently on the LiDAR point cloud [33].…”

Section: Target-based Approachmentioning

confidence: 99%

CFNet: LiDAR-Camera Registration Using Calibration Flow Network

Wang

Dou

et al. 2021

Sensors

View full text Add to dashboard Cite

show abstract

“…In the scanning coordinate system, the scanning coordinate (X, Y, Z) of measuring point p was affected not only by spherical coordinates but also by the position of spherical coordinates (x 0 , y 0 , z 0 ) in the scanning coordinate system. At this point, the scanning coordinate of measuring point p should be calculated by Equation (7). The sphere target point cloud was simulated by equally dividing the zenith angle θ and plane projection angle ϕ.…”

Section: Point Cloud Simulationmentioning

confidence: 99%

“…Part of its contour information could be obtained from any angle of view, with which the spherical center and radius could be effectively solved. Therefore, it was widely used in the multi-class application research of terrestrial LiDAR, such as the calibration and check of a terrestrial laser scanner, scanning accuracy evaluation, registration, and georeferencing of point clouds et al [6][7][8][9][10][11]. The geometric center of the sphere target was inside the sphere and could not be obtained directly through measurement.…”

Section: Introductionmentioning

confidence: 99%

An Algorithm for Fitting Sphere Target of Terrestrial LiDAR

Shi

Zhao

Wang

et al. 2021

Sensors

View full text Add to dashboard Cite

The sphere target played a vital role in terrestrial LiDAR applications, and solving its geometrical center based on point cloud was a widely concerned problem. In this study, we proposed a newly finite random search algorithm for sphere target fitting. Based on the point cloud data and the geometric characteristics of the sphere target, the algorithm realized the target sphere fitting from the perspective of probability and statistics with the help of parameter estimation. Firstly, an initial constraint space was constructed, and the initial center and radius were determined by finite random search. Then, the optimal spherical center and radius were determined gradually through continuous iterative optimization. We tested the algorithm with the simulated and realistic point cloud. Experimental results showed that the proposed algorithm could be effectively applied to all kinds of point cloud fitting. When the coverage rate was bigger than 30%, the fitting accuracy could reach within 0.01 mm for all kinds of point clouds. When the coverage rate was less than 20%, the fitting accuracy can reach ±1 mm, although it was reduced to a certain extent.

show abstract

“…Geiger et al (12) used the idea of classifying the calibration plate laser point cloud and eliminating the point sets with a significantly low number of 3D points to achieve alignment. Hand et al (13) and Li et al (14) used LIDAR data points lying on a straight line to calibrate constraint relations using isosceles triangles and folded flat panels as calibrators, respectively. Guindel et al (15) achieved the automatic calibration of LIDAR and a stereo camera by using simple calibrators.…”

Section: Introductionmentioning

confidence: 99%