Section: New Results

Motion, Scene and Camera Reconstruction

Participants : Marie-Odile Berger, Srikrishna Bhat, Pierre Rolin, Gilles Simon, Frédéric Sur.

• Metrological performance enhancement and resolution assessment for experimental solid mechanics

  This work is motivated by image processing problems from experimental solid mechanics. One of the problem in this field is to measure heterogeneous strains on the surface of specimens subjected to mechanical tests, through an imaging device. Among full-field measurement techniques, the grid method consists in transferring a regular grid on the surface of the specimen and in taking images of the grid before and after deformation. Windowed Fourier analysis then gives an estimation of the surface displacement and strain components. In a collaboration with Institut Pascal (Université Blaise Pascal, Clermont Ferrand), we have shown that the estimations obtained by this technique are a first-order approximation of the convolution of the actual values with the analysis window. We have also characterized how the noise in the grid image impairs the displacement and strain maps (F. Sur, M. Grédiac. Towards deconvolution to enhance the grid method for in-plane strain measurement. To appear in AIMS Inverse Problems and Imaging, American Institute of Mathematical Sciences, 2014.). This study has allowed us to improve the metrological performance of the grid method with deconvolution algorithms. A numerical and experimental study can be found in [10] , [16] , [21] . As any contactless measurement method, the resolution of the grid method is limited by the noise impairing the sensor. We have also characterized this resolution within a Poisson-Gaussian noise model, which is known to be realistic for CCD or CMOS sensors  (M. Grédiac, F. Sur. Effect of sensor noise on the resolution and spatial resolution of displacement and strain maps estimated with the grid method. To appear in Strain, Wiley, 2014.).

• Matching in difficult conditions

  Visual vocabularies are emerging as a new tool for building point correspondences for pose estimation. Within S. Bhat’s PhD thesis [9] we have proposed several methods for visual word construction dedicated to point matching, with structure from motion and pose estimation applications in view. The three dimensional geometry of a scene is first extracted with bundle adjustment techniques based on keypoint correspondences. These correspondences are obtained by grouping the set of all SIFT descriptors from the training images into visual words. We obtain a more accurate 3D geometry than with classical image-to-image point matching. In a second on-line step, these visual words serve as 3D point descriptors that are robust to viewpoint change, and are used for building 2D- 3D correspondences on-line during application, yielding the pose of the camera by solving the PnP problem. Several visual word formation techniques have been compared with respect to robustness to viewpoint change between learning and the test images.

  The PhD thesis of P. Rolin comes within the scope of camera pose estimation from an unstructured 3D point dataset, endowed with image descriptors. His work focuses on improving pose estimation with respect to strong viewpoint changes. 2D-3D correspondences are actually difficult to establish if there are too large viewpoint changes between the image whose pose is sought and the images that gave the 3D point dataset. P. Rolin currently assesses viewpoint simulation techniques in order to enhance the description of the 3D points with information from different viewpoints.

• Acquisition of 3D calibrated data

  In situ modeling is generating increasing interest in the community as it makes it possible to build AR applications in unprepared environments. In [19] , we present a new method for interactive modeling of polygonal scenes, using a tablet PC, a laser rangefinder, an inertial measurement unit (IMU) and a camera. A well-founded calibration method is used to determine the orientation of the IMU and the origin and direction of the laser beam in the camera coordinate system. A new hybrid, driftless orientation tracking method is proposed, inspired by the tracking-by-synthesis algorithm adapted to 3-degree-of-freedom camera motions. Visual hints are provided during the tracking-and-modeling process in order to help the user get the best possible accuracy. These visual hints are based on a PCA analysis of the reconstructed laser point clouds and statistical measurements of the camera tracking accuracy.