B401 ニューラルネットワークを用いた勾配法に基づくPIV計測

Abstract

The gradient-based method, in which an optical flow (a velocity vector profile) is obtained from spatial and temporal image gradients, has been often applied to not only PIV but also Robot Vision just because it has a high spatial resolution. In the method, a two-dimensional velocity vector profile is measured by solving an basic equation based on the relationship between two velocity vector components and the image gradients. The method, however, needs another restraint condition because two unknown velocity components are not obtained by solving just one equation. We propose a new method using an artificial neural network for learning vector fields by which the above problem is solved. The neural network is trained by using spatial and temporal image gradients as teaching data so that the basic equation for the gradient method is satisfied. As the neural network itself learns the stream function, the continuity equation of flow is consequently satisfied in the measured velocity vector fields.