离焦数字粒子图像测速中粒子中心检测方法

摘要

在离焦数字粒子图像测速(Defocusing digital particle image velocimetry,DDPIV)技术中采用2D高斯函数作为离焦粒子灰度分布模型具有一定的局限性.当粒子模糊半径较大时,难以有效地检测出粒子中心位置.针对此问题,基于圆的霍夫变换(Circular Hough transform,CHT)提出了一种新的粒子中心检测方法,有效克服了原有方法中存在的缺陷.首先,将原始粒子图像分割为多个具有部分重叠的子粒子图像,并对子粒子图像中粒子的边缘进行检测和细化.其次,对获得的粒子边缘进行CHT变换,利用CHT变换空间的峰值特性和重叠粒子分层检测方法可获取无重叠粒子、遮挡粒子和被遮挡粒子的中心位置.最后,对本文方法分别进行了仿真和实验.实验结果表明,本文方法具有较高的精度和较好的鲁棒性.%In defocusing digital particle image velocimetry (DDPIV), it is difficult to detect the particle centroid since the commonly-used 2D Gaussian function cannot fit the defocused particle intensity profile accurately. Aiming at the issue, a new particle centroid detection method,based on the circular Hough transform, is proposed. Firstly, the original particle image is divided into several overlapping sub-regions, and the particle edges are detected and thinned.Secondly, the circular Hough transform and the hierarchical particle detection method used to detect the centroids of non-overlapped particle, overlapped particle and overlapping particle,respectively. Finally, the proposed method is tested by simulation experiments and real experiments. Experimental results demonstrate the accuracy and robustness of the proposed method.