Section: New Results

Parameter estimation

Metrologic performance assessment in experimental mechanics

A problem of interest in experimental solid mechanics is strain map estimation on the surface of a specimen subjected to a load or a tensile test. One of the available approaches is based on images of a pseudo-periodic grid transfered on the surface of the specimen. Sensor noise is a major source of uncertainty in the strain map, and quantifying the propagation of the sensor noise to the measured strain components is a major problem when metrological performances are in view. We have proposed in [12] a study of the mathematical properties of the popular method based on windowed Fourier analysis, under a Gaussian white noise assumption. In the case of a more realistic signal-dependent, heteroscedastic noise, we have quantified in [10] (see also [15] , [26] ) the trade-off between the noise amplitude, the measurement resolution and the spatial resolution of the method. We have also investigated image stacking for noise reduction. While averaging a serie of images is certainly the most basic option to reduce the noise, it is not effective for studying grid images under a high magnification factor, because of unavoidable residual vibrations carried for instance by concrete floor slabs. We have shown in [13] that, while these vibrations indeed blur grid images, they still permit to reduce the noise amplitude in the displacement and strain maps.

Sensor noise measurement.

While searching for a low-cost on-the-fly estimation of the sensor parameters based on a serie of grid images (thus with no need of changing the experimental setting), we have proposed in [11] an algorithm which is able to deal with the vibrations biasing the estimations. More generally, we have investigated in [21] the problem of sensor parameter estimation from a series of images, under light flickering and vibrations. Light flickering is indeed a natural assumption for indoor artificial lights. It is also involved by slight variations in the opening time of a mechanical shutter. We have proposed a model of the pixel intensity based on a Cox process, together with an algorithm which, taking benefit of flickering, gives an estimation of every sensor parameter, namely the gain, the readout noise, and the offset.

Image driven simulation

In the IDeaS ANR project we propose to target Image-driven simulation, applied to interventional neuroradiology: a coupled system of interactive computer-based simulation (interventional devices in blood vessels) and on-line medical image acquisitions (X-ray fluoroscopy). The main idea is to use the live X-ray images as references to continuously refine the parameters used to simulate the blood vessels and the interventional devices (micro-guide, micro-catheter, coil).

Our guideline is to follow a sequential statistical filtering approach to fuse such heterogeneous data. This approach first calls for an improved knowledge of the statistical behavior of the simulation, which we addressed in the past year through experimental studies. We described our experimental setup in [20] , which, in particular uses high speed stereo reconstruction to be able to study non quasi-static effects. Preliminary measures of the catheter speed during stick and slip transitions back up our conviction that quasi-static mechanical models fail to simulate such rapid motions of the tool. Our on-going analysis of the simulation sensitivity to mechanical parameters also sets forward friction as critical for high-fidelity simulation.