Fig 3 - uploaded by Shengdong Xu
Content may be subject to copyright.
Source publication
In this paper, we develop a way to accurately and precisely estimate the pose of a calibrated camera with a single picture which includes a known planar object. For the proposed algorithm, we first use SURF detector for feature extraction and matching. Then, we use the information from known reference image to retrieve 3D point coordinates. Based o...
Context in source publication
Context 1
... the object is planar, if we reconstruct 3D points of the features, most of them should be coplanar. Here, we use triangulation method to reconstruct 3D points coordinates based on the information from both query image and reference image. The algorithm is shown as Figure 3. Using classical absolute pose estimation matrix, we can get an initial result (rotation matrix R i and translation vector T i from the world frame to query image camera frame). Given that the reference pose is known as a priori (rotation matrix R r and translation vector T r from world frame to reference image camera frame are known), the rotation matrix R and translation vector T from the reference image camera frame to query image camera frame can be easily ...
Citations
... Techniques that solely rely on monocular geometry necessitate a controlled environment and additional reference information, making them more application-specific [19][20][21][22]. Nevertheless, monocular techniques offer a cost-effective solution in that they require fewer devices, and are still quite accurate [12]. ...
Automatic measurements via image processing can accelerate measurements and provide comprehensive evaluations of mechanical parts. This paper presents a comprehensive approach to automating evaluations of planar dimensions in mechanical parts, providing significant advancements in terms of cost-effectiveness, accuracy, and repeatability. The methodology employed in this study utilizes a configuration comprising commonly available products in the industrial computer vision market, therefore enabling precise determinations of external contour specifications for mechanical components. Furthermore, it presents a functional prototype for making planar measurements by incorporating an improved subpixel edge-detection method, thus ensuring precise image-based measurements. The article highlights key concepts, describes the measurement procedures, and provides comparisons and traceability tests as a proof of concept for the system. The results show that this vision system did achieve suitable precision, with a mean error of 0.008 mm and a standard deviation of 0.0063 mm, when measuring gauge blocks of varying lengths at different heights. Moreover, when evaluating a circular sample, the system resulted in a maximum deviation of 0.013 mm, compared to an alternative calibrated measurement machine. In conclusion, the prototype validates the methods for planar dimension evaluations, highlighting the potential for enhancing manual measurements, while also maintaining accessibility. The presented system expands the possibilities of machine vision in manufacturing, especially in cases where the cost or agility of current systems is limited.
... Pose estimation of geometric elements is a common requirement in 2D and 3D vision based applications, where different methods have been developed and studied to tackle this problem from different point of views. In the bibliography different methods (see figure 1) can be found such as pose orientation from known points (Oberkampf et al 1996, Triggs 1999, Zhi and Tang 2002, Xu and Liu 2013, pose estimation using shape-based 3D matching (Osada et al 2001, Rosenhahn et al 2006, Teck et al 2010, Yang et al 2016, pose estimation using surface-based 3D matching (Rabbani and Van Den Heuvel 2005, Su and Bethel 2010, Paláncz et al 2016, Figueiredo et al 2017 and pose estimation based on model-based approaches. Pose estimation is required in motion applications as well as for measuring, binpicking or even alignment tasks. ...
In this research, a multi-view photogrammetric model was developed and tested in simulation in order to understand its capabilities for close-range photogrammetric applications. It was based on contour line detection and least squares geometrical fitting of a cylindrical geometry from multiple views. To feed and validate this model, synthetic data were created for several cylinders poses and camera network set-up. The simulation chain comprises three main stages: synthetic image creation, image data processing by means of shape-matching and cylinder pose estimation based on developed photogrammetric model. Beforehand, a priori data was theoretically established according to a common reference for both for intrinsic and extrinsic parameters of the cameras. The preliminary results highlight that the model is suitable for close-range photogrammetry and sensible to a priori known data as well as to image data quality. These results were compared against other validated geometrical methods to assure that the model is truthful. Preliminary results show that the accuracy of the model ranges between 1/1000 and 1/20 000 for multiple poses and cylinder dimensions. Moreover the simulation procedure has been enhanced with a Montecarlo approach to estimate more realistic pose uncertainties considering possible imaging error sources.
... The research on the pose estimation issue involves a wide range of approaches and algorithms. Many descriptors, such as PFH and FPFH [25,32,35] [31]. However, the points generated by 3-D sensor are not continuously defined in the 3-D space. ...
In this paper, we deal with the problem of pose estimation based on point cloud. We modify the Iterative closest face (ICF) algorithm by mathematical techniques, in which a new method to calculate point-face distance with less computational cost is proposed. Then, we combine this algorithm with particle swarm optimization to get a better searched result. PSO is employed because there are few parameters to adjust and it is more efficient than the original searched method in ICF. A set of experiments is conducted, following the statistical analysis of the results. These experiments demonstrate the accuracy and robustness of our algorithm.
