Stereo-LFC PIV and Stereo-Multigrid enhanced with image distortion: Study on the limits of resolvable scales (original) (raw)
Related papers
Combination of advanced 2D PIV and Stereo Technique
… on application of …, 2006
Particle Image Velocimetry with Local Field Correction (LFC-PIV) has been tested in the past to obtain two components of the velocity in a two dimensional domain (2D2C). When compared with conventional correlation based algorithms, this advanced technique has showed improvements in three important aspects: robustness, resolution and ability to cope with large displacements gradients. A further step in the development of PIV algorithms consists in the combination of LFC with the Stereo technique, able to obtain three components of the velocity (2D3C PIV). To successfully apply stereoscopy, the laser sheet width is usually larger than in 2C velocimetry, in order to acquire the out of plane displacement of the particles. This peculiarity can make questionable the need for improved spatial resolution. In particular, if some factors concur the measurement of small scale features of the velocity along the laser sheet can be spoiled by the velocity variations across it, caused by the flow structures. Being the spatial resolution capabilities of LFC its main advantage, this paper focuses in establishing if there are advantages left for LFC-PIV in situations where only large scale features are of interest, thus making it difficult for this technique to show its attractiveness. The remaining benefits still to be analyzed are the robustness and the ability to cope with large displacement gradients. For both aspects, the performances of 2D3C-LFC-PIV are explored. These performances are characterized and compared with conventional and other advanced algorithms, when applied to synthetic PIV images. In addition, the coherence between these results and those coming from experimental PIV images is presented and discussed.
One-step Stereo PIV: theory and performances
The purpose of this communication is to theoretically analyze and assess the performances of a recently proposed novel Stereo PIV computation approach (Leclaire et al. 2009). The specificity of the algorithm is that it computes the three-component displacement corresponding to an interrogation window directly in one step, whereas conventional algorithms involve two steps, first interrogating separately image pairs corresponding to each camera, and then obtaining the final displacement as a result of the so-called stereo reconstruction. We first analyze theoretically the differences between the two approaches. The direct approach in fact amounts to seek the displacements pairs directly in the admissible subspace induced by the stereoscopic geometry, automatically excluding possible spurious pairs, whereas in the two-step case, peak pairs are constituted by matching the sharpest peak of each image, and then projected onto the closest pair in the subspace. We then show that under ideal imaging conditions, both approaches are theoretically equivalent, provided that the correlation peaks of each image have the same curvature. We then assess the sensitivity of both approaches to various relevant parameters, using synthetic images. The tests performed confirm that identical results are obtained for perfect imaging conditions. They also show that while the onestep approach is slightly more sensitive to peak-locking, and in a more pronounced way to the registration error, significant gains are observed in situations with low seeding, loss of particle pairs and low signal-tonoise ratio. Highest gains are achieved in particular when cameras are characterized by a different signal-tonoise ratio, which may occur in particular in forward/backward diffusion setups, whereby the method automatically weighs more importantly the data with the highest signal to noise ratio.
Target-free stereo PIV: a novel technique with inherent error estimation and improved accuracy
2008
Abstract A novel, accurate and simple stereo particle image velocimetry (SPIV) technique utilising three cameras is presented. The key feature of the new technique is that there is no need of a separate calibration phase. The calibration data are measured concurrently with the PIV data by a third paraxial camera. This has the benefit of improving ease of use and reducing the time taken to obtain data. This third camera also provides useful velocity information, considerably improving the accuracy of the resolved 3D vectors.
Advances in computational stereo
2003
Abstract Extraction of three-dimensional structure of a scene from stereo images is a problem that has been studied by the computer vision community for decades. Early work focused on the fundamentals of image correspondence and stereo geometry. Stereo research has matured significantly throughout the years and many advances in computational stereo continue to be made, allowing stereo to be applied to new and more demanding problems.
Proceedings of International Symposium on Computer Vision - ISCV, 1995
One of the most difficult aspects of developing computational algorithms for stereopsis that match the intrinsic capabilities of human vision is the correspondence problem; that is locating the same point, if it exists, in multi-viewed time-varying sensor measurements. Correspondences have been determined using feature-based or region-based matching algorithms with bottom-up or top-down implementations [3]. The bottom-up or low-level approach for stereo analysis includes: i) extracting feature points or area measures in both views, ii) matching the feature points or area measures under certain geometric, illumination, reflectance and object constraints, and iii) computing a depth or height map using the disparity values from correspondences using sensor geometry and scanning configuration. Most stereo algorithms invariably produce errors due to noise, low image or feature content, geometric distortion, depth discontinuities, occlusion, illumination and reflectance changes across the scene and between views, transparency effects leading to multiple matches, and instability of the cameras and sensors during image formation. Such model violations are difficult to handle in a comprehensive fashion. Robust statistical methods have recently been applied to a variety of computer vision problems including motion estimation [lo] [ll][S], surface recovery from range data [9], and image segmentation [2]. Robust methods offer a powerful alternative to smoothness and regularization constraints to mitigate the effects of model errors. A new multistage adaptive robust (MAR) algorithm combined with a multiresolution coarse-to-fine matching model is developed for robust stereo analysis.
A computational framework for stereo imaging
2005
Stereo vision recovers depth information from a 3-D scene by exploiting differences between two different 2-D projections. This involves two processes: analysis and reconstruction. The former estimates camera parameters-calibration-and obtains the relative position of the same 3-D point in the two different images-correspondence estimation. The latter deals with the extraction of the 3-D scene information. These two ill-posed problems are the focus of research in stereo imaging. viii ______________________________________________________________________
Turbulent flow height measurement with stereo vision
This paper describes the process of 3D analysis of two water currents with method of photogrammetry. Photogrammetry is used in fields such as architecture, engineering, police investigation, preserving cultural heritage, military and geology. This method can be used in military to reconstruct a site with traces of shrapnel or various projectiles. In our case we tried to measure height of turbulent flow, where two currents collided at the angle of 90°. In first section we introduce our problem and method. Second section describes method of photogrammetry and basics of torrential flows. In third section we describe our experiment. Fourth section describes the course of getting 3D model. In fifth section we analyze results and in sixth section, a conclusion is given.
Verification of 3-D Stereoscopic PIV Operation and Procedures
Particle image velocimetry (PIV) is a nonintrusive, whole-field velocity measurement technique that has been used since the mid-1980s. The accuracy, flexibility and versatility offered by PIV systems have made them extremely valuable tools for flow studies. 3-D stereoscopic PIV is the package capable of measuring 3-dimensional velocity components. It involves a very sophisticated routine during setup, calibration, measurement and data processing phases. This paper aims to verify the procedures of operation used for 3-D stereoscopic PIV measurements. This is important to ensure that the best data representation with low associated uncertainty is obtained. A free-diffuser inlet of rectangular cross-section, 14.2 cm x 6.2 cm, with known local air velocities (i.e. measured using pitot-static probe), is presently considered. The flow is assumed to be fully developed turbulent since sufficient hydrodynamic entry length, 4.4D h Re 1/6 < L h,turb < 50D h and Reynolds Number, Re>10000 are introduced. Images that are captured by CCD cameras are interpreted using Dantec Dynamic software providing 3-dimensional velocity vectors. The velocities obtained from PIV and pitot-static probe are compared in order to justify the quality of PIV measurement. The range of velocity obtained using probe is 2.31-2.58 m/s, whereas using PIV is 2.31-2.91 m/s. It thus gives the average discrepancy of 0.8%. Besides, there is also a close agreement between the air velocities measured by PIV and theories with average discrepancy of 1.2%. This discrepancy is mainly due to some uncertainties in the experiments such as imperfect matching of coordinate between probe and laser sheet, unsteadiness of flow, variation in density and less precision in calibration. The operating procedures of 3-D stereoscopic PIV have successfully been verified thus are justified to be used for future PIV measurement, provided minor discrepancies are recorded. Index Term-3-D stereoscopic particle image velocimetry (PIV), uncertainty analysis (UA).
EVALUATION OF TWO RECONSTRUCTION SYSTEMS BASED ON STEREOVISION
researchgate.net
This paper presents an evaluation and comparison of two 3D reconstruction approaches. The first one is an existing sparse reconstruction which is based on the triangulation of extracted points of interest using calibration parameters. The results obtained by this approach present the outline of the object. The second one is improved dense reconstruction approach based on a generated disparity map. For this approach, we improve the stereo matching algorithm. In fact, the initial disparity map is computed using a weighted dissimilarity measure with application of adaptive correlation windows. Then, planes are modeled by a set of planar surface patches. First, an initial set of planes is estimated using only reliable pixels. Then, those planes will be merged to obtain a final representation of the scene. Thus, each pixel and region will be assigned to a disparity layer according to a cost function which is subject of minimization. The experimentations and evaluations show that the second approach gives efficient results concerning the recovered informations.
The CMP Evaluation of Stereo Algorithms
2003
Detailed studying of stereo matching algorithm properties and behaviour under varying conditions is crucial for the algorithm improvement and development. Towards this purpose we have designed a ground-truth evaluation method focusing on algorithm failure due to insufficient signal-to-noise ratio, since the data uncertainty is never totally avoidable. For the complex evaluation, nine types of error statistics have been defined. The errors are focussed on basic matching failure mechanisms and their definitions observe the principles of orthogonality, symmetry, completeness and independence. The test scene consists of a background plane and five thin stripes on a foreground. It has been captured under 20 different levels of texture contrast, in each level 10× with randomly shifted texture. The scene has been designed to preserve ordering. The ground truth has been obtained semi-manually. The algorithms are evaluated on all the 200 images and the results are shown as disparity maps and...