功率定律和逆运动学建模:应用于从卫星图像测量湍流
In the context of tackling the ill-posed inverse problem of motion estimation fromimage sequences, we propose to introduce prior knowledge on ow regularity given byturbulence statistical models. Prior regularity is formalized using turbulence powerlaws describing statistically self-similar structure of motion increments across scales.The motion estimation method minimizes the error of an image observation modelwhile constraining second order structure function to behave as a power law within aprescribed range. Thanks to a Bayesian modeling framework, the motion estimationmethod is able to jointly infer the most likely power law directly from image data. Themethod is assessed on velocity elds of 2D or quasi-2D ows. Estimation accuracyis rst evaluated on a synthetic image sequence of homogeneous and isotropic 2Dturbulence. Results obtained with the approach based on physics of uids outperformsstate-of-the-art. Then, the method analyzes atmospheric turbulence using a realmeteorological image sequence. Selecting the most likely power law model enables therecovery of physical quantities which are of major interest for turbulence atmosphericcharacterization. In particular, from meteorological images we are able to estimateenergy and enstrophy uxes of turbulent cascades, which are in agreement withprevious in situ measurements.