EKFPnP: extended Kalman filter for camera pose estimation in a sequence of images

Abstract

In real-world applications the perspective-n-point (PnP) problem should generally be applied to a sequence of images which a set of drift-prone features are tracked over time. In this study, the authors consider both the temporal dependency of camera poses and the uncertainty of features for the vision-only sequential camera pose estimation. Using the extended Kalman filter (EKF), a priori estimate of the camera pose is calculated from the camera motion model and then it is corrected by minimising the reprojection error of the reference points. Applying probabilistic approach also provides the covariance of the pose parameters which helps to measure the reliability of the estimated parameters. Experimental results, using both synthetic and real data, demonstrate that the proposed method improves the robustness of the camera pose estimation, in the presence of tracking error and feature matching outliers, compared to the state of the art.

Flexible High-Resolution Continuous 3-D Scanning for Large-Scale Industrial Components

“…Therefore, further pose processing is needed to suppress the fluctuation and drift between neighboring poses that significantly affect the shape of the local point cloud features after registration. For pose estimation based on an area-scan camera, we suppress the fluctuation by considering the temporal dependency of poses through the image sequence, and the EKF-based PnP (EKFPnP) [44] is applied. Appendix A provides details of the EKFPnP.…”

A High-Accuracy, Scalable and Affordable Indoor Positioning System Using Visible Light Positioning for Automated Guided Vehicles

“…In contrast to the custom algorithms presented in [17,18], the use of this solvePnP presents some limitations in terms of the number of minimum references necessary to obtain a position, needing either four references at the same height (or plane), or six references. However, as it is a tool with good support and it is easy to use, it is becoming a preferred option for obtaining positions in 3D from a 2D image [19][20][21][22][23].…”

Gated Region-Refine pose transformer for human pose estimation

“…Using MATLAB, we tested our algorithm against existing PnP algorithms calculating 3D pose for general and planar configurations of points. For the general case, points were randomly generated in the cuboid [−2, 2] × [−2, 2] × [4,8] in camera space, projected onto the image plane assuming a focal length of 800 pixels, and each image point was perturbed with Gaussian noise. This follows the methodology of [9], and the implementations that we tested came from their MATLAB toolbox.…”

“…Some algorithms, such as IPPE, attempt to compute all possible solutions, either selecting the best fit or returning the solutions to the user. The outputs of existing algorithms can be filtered as in [8]. However, as robust as these solutions are, they solve only a facet of the deeper issue of imprecision for planar configurations.…”