Estimation of Centers and Stagnation points in optical flow fields

In a topological sense fluid flows are characterised by their stagnation points. Given a temporal sequence of images of fluids we will consider the application of local polynomials to the estimation of smooth fluid flow fields. The normal flow at intensity contours is estimated from the local distribution of spatio-temporal energy, which is sampled using a set of spatio-temporal quadrature filters. These observations of normal flows are then integrated into smooth flow fields by locally approximating first order polynomials in the spatial coordinates to the flow vectors. This technique furthermore allows us to give a qualitative local description of the flow field and to estimate the position of stagnation points (e.g. nodes, saddles, and centers). We will apply the algorithm to two data sets. The first sequence consists of infrared images from the meteorological satellite Meteosat. Here the purpose is that of estimating cloud motion. The second sequence visualises the airflow in a model of a livestock building by inducing smoke in the air inlets and illuminating a plane using a laser sheet. In this case the task is to estimate the flow field in order to evaluate the ventilation system.