Adaptive vector validation in image velocimetry to minimise the influence of outlier clusters
Adaptive vector validation in image velocimetry to minimise the influence of outlier clusters
复制标题
DOI：
10.1007/s00348-015-2110-8
复制
发表时间：
2016-02
影响因子：
2.4
通讯作者：
A. Masullo;R. Theunissen
中科院分区：
文献类型：
--
作者：
A. Masullo;R. Theunissen
通用离群值检测方案（Westerweel和Scarano in Exp Fluids 39:1096-1100, 2005）和非结构化数据的距离加权通用离群值检测方案（Duncan et al. in Meas Sci technology 21:05 - 7002, 2010）是最常见的PIV数据验证例程。然而，这些技术依赖于每个向量与固定大小的邻居的空间比较，并且它们的性能随后在异常值集群的存在下受到影响。本文提出了一种改进方法，使异常点检测更加鲁棒，同时降低错误地使正确向量无效的概率。速度场在局部相干性方面进行初步评估，这参数化了相邻区域的范围，随后将与每个矢量进行比较。这种自适应被证明可以减少未检测到的异常值的数量，即使在上述验证方案中实现也是如此。此外，作者提出了一种考虑矢量大小和角度的替代残差定义，采用改进的基于高斯加权距离的平均中位数。该程序能够使可接受的速度背景波动程度适应于局部位移幅度。基于孤立涡流场、湍流通道流场和强迫各向同性湍流的DNS模拟，对传统的、扩展的和推荐的验证方法进行了数值评估。由此产生的验证方法是自适应的，不需要用户定义的参数，并被证明在异常值检测不足和过度方面产生最佳性能。最后，将新的验证程序应用于多孔盘后近尾迹和超音速射流实验研究的PIV分析，说明了在空间分辨率和精度方面的潜在收益。
The universal outlier detection scheme (Westerweel and Scarano in Exp Fluids 39:1096–1100, 2005) and the distance-weighted universal outlier detection scheme for unstructured data (Duncan et al. in Meas Sci Technol 21:057002, 2010) are the most common PIV data validation routines. However, such techniques rely on a spatial comparison of each vector with those in a fixed-size neighbourhood and their performance subsequently suffers in the presence of clusters of outliers. This paper proposes an advancement to render outlier detection more robust while reducing the probability of mistakenly invalidating correct vectors. Velocity fields undergo a preliminary evaluation in terms of local coherency, which parametrises the extent of the neighbourhood with which each vector will be compared subsequently. Such adaptivity is shown to reduce the number of undetected outliers, even when implemented in the afore validation schemes. In addition, the authors present an alternative residual definition considering vector magnitude and angle adopting a modified Gaussian-weighted distance-based averaging median. This procedure is able to adapt the degree of acceptable background fluctuations in velocity to the local displacement magnitude. The traditional, extended and recommended validation methods are numerically assessed on the basis of flow fields from an isolated vortex, a turbulent channel flow and a DNS simulation of forced isotropic turbulence. The resulting validation method is adaptive, requires no user-defined parameters and is demonstrated to yield the best performances in terms of outlier under- and over-detection. Finally, the novel validation routine is applied to the PIV analysis of experimental studies focused on the near wake behind a porous disc and on a supersonic jet, illustrating the potential gains in spatial resolution and accuracy.