Application of Particle Image Velocimetry Method Based on Full Convolution Neural Network in Turbulence Field

Exploring fluid flow phenomena has always been a hot topic in fluid mechanics research. Particle Image Velocimetry (PIV) is a non-contact technique applied in fluid mechanics laboratories for measuring the velocity field of fluids. The core of PIV lies in extracting displacement vector information from particle image pairs to ultimately generate a two-dimensional velocity field. Traditional methods typically estimate the two-dimensional velocity field using cross-correlation algorithms and optical flow techniques. In recent years, with the rapid development of artificial intelligence technologies and their high compatibility with PIV in image processing, there is significant potential and research value in designing deep learning techniques for accurately estimating fluid velocity vector fields in particle image velocimetry. Compared to traditional methods, the PIV technique proposed in this paper, based on fully convolutional neural networks, not only maintains the accuracy of velocity field estimation but also significantly enhances resolution and computational efficiency. Experiments demonstrate that the proposed method exhibits higher accuracy and stronger generalization capabilities.