Whole Field Measurement by Gradient-Based PIV
I. Kimura, Y. Ono, M. Ohta, A. Nakata, A. Kaga and Y Kuroe
We have proposed a novel gradient-based PIV using an artificial neural network for acquiring the characteristics of two-dimensional flow fields. The neural network which outputs the stream function is trained by using spatial and temporal image gradients so that the basic equation of the gradient method is satisfied. The gradient-based PIV can consequently realize an accurate approximation of two-dimensional flow fields by using those image gradients. In this paper the proposed PIV is applied to both artificially-generated deficient smoke images and experimentally-visualized tracer images. The former examination shows that the method makes the whole field measurement feasible even from such deficient images. The latter one proves that even velocity vectors very close to a wall, which are unmeasurable by conventional PIV, are measurable.