AI Summary & Key Takeaways
AI InsightResults show that histogram thresholding-based masking provides better results for swimmers which randomly change shape and direction; whereas, feature tracking-based masking provides better results for swimmers which do not change shape or direction significantly.
📌 3 Key Takeaways
Özet
Objects and surfaces often appear in Particle Image Velocimetry (PIV) images. Unless masked, the features on these contribute to the cross correlation function and introduce an error in the vectors as a result of the PIV analysis in the vicinity of the phase boundary. Digital masking of objects has appeared numerous times in the literature as part of the analysis chain, with a growing focus on isolating moving features using dynamic masks. One aim of this article is to provide a summary of milestones achieved in dynamic masking covering a wide range of applications. Another aim is to show the difference between image masking and vector masking. Finally, two different dynamic masking examples are described in detail and compared. The examples used are selected from swimming microorganisms in small channels. In the first example, a histogram thresholding-based dynamic masking is used, while, in the second example, a novel technique employing a feature tracking-based dynamic masking is used. Results show that histogram thresholding-based masking provides better results for swimmers which randomly change shape and direction; whereas, feature tracking-based masking provides better results for swimmers which do not change shape or direction significantly. In order to show the improvement due to dynamic masking, a comparison is made between PIV results a) with no masking, b) with just image masking and c) with both image and vector masking. Results show that the best approach is to use both image and vector masking.