DEVELOPMENT OF HIGH-ACCURATE PARTICLE IMAGE VELOCIMETRY BASED ON IMPROVED OPTICAL FLOW METHOD

摘要

A new method of image analysis for measuring visualized two-dimensional flow field was developed. The method is basically based on the optical flow technique which is one of the image analysis techniques for quickly detecting moving objects, e. g. in the field of robotics. Conventionally optical flow technique utilizes an advection equation for image intensity field by assuming that local velocity is kept constant within a short time interval. In the new method, constant velocity components as well as velocity gradients are taken into account to improve the accuracy in measuring shearing flow fields. The performance of the proposed method was evaluated by using synthetic images with various velocity distributions and also actual surface flow images obtained during flood. The presented method has several significant advantages over the conventional PIV (Particle Image Velocimetry).