Particle recognition and particle matching are the core of particle tracking velocimetry (PTV). Particle recognition and matching directly determine the accuracy of flow field analysis, and particle recognition is a prerequisite for matching. In the engineering application of PTV technology, there are various reasons for the image noise generated by shooting, which seriously interferes with particle recognition. Accurately separating particles from noise is the most important and difficult process in particle recognition while ensuring that particle information is not damaged. According to the particle image captured by PTV experiment, the author has proposed a new processing method for denoising and recognition of noise-containing original image (ImageDenoising-ParticleRecognition). This method comprehensively uses SUSAN detection, expansion calculation, threshold segmentation, four-connected mark, non-particle culling, particle hole filling and other technologies to protect the image information of particles from loss to the maximum extent and ensure that the noise-containing particle images can complete particle recognition in one step. At the same time, the method proposed in this paper more accurately determines the edge of the particle image, and provides a more reliable particle image for the calculation of the flow field motion in the later stage. For the method proposed in this paper, based on the Visual C++ platform autonomous programming, the particle images generated by computer simulation and the SiO₂ particle image with 500 nm diameter shot by the actual PTV experiment have been verified and analyzed respectively, and the recognition results with good accuracy have been obtained.

Graphic abstract

The preprocessing of particle image is the prerequisite of PTV processing, and its level directly has a crucial impact on the final analysis accuracy of the flow field. According to the particle image captured by PTV experiment, the author has proposed a new preprocessing method to provide a more reliable particle image for the calculation of the flow field motion in the later stage.

中文翻译：

---

PTV图像的噪声处理与粒子识别算法研究

粒子识别和粒子匹配是粒子跟踪测速（PTV）的核心。颗粒识别和匹配直接决定了流场分析的准确性，而颗粒识别是匹配的前提。在PTV技术的工程应用中，由于拍摄而产生的图像噪声有多种原因，严重干扰了颗粒识别。在确保粒子信息不被破坏的同时，准确地将粒子与噪声分离是最重要和最困难的过程。根据PTV实验采集的粒子图像，提出了一种对含噪声原始图像进行降噪和识别的新方法（ImageDenoising-ParticleRecognition）。此方法全面使用了SUSAN检测，扩展计算，阈值分割，四连通标记，非粒子剔除，粒子孔填充等技术，最大程度地保护粒子的图像信息不丢失，并确保含噪声的粒子图像可以一次完成粒子识别步。同时，本文提出的方法可以更准确地确定粒子图像的边缘，并为以后阶段的流场运动计算提供更可靠的粒子图像。对于本文提出的方法，基于Visual C ++平台自主编程，通过计算机模拟和SiO生成粒子图像。粒子孔填充和其他技术可最大程度地保护粒子的图像信息免受损失，并确保包含噪声的粒子图像可以一步完成粒子识别。同时，本文提出的方法更准确地确定了粒子图像的边缘，并为后期的流场运动计算提供了更可靠的粒子图像。对于本文提出的方法，基于Visual C ++平台自主编程，通过计算机模拟和SiO生成粒子图像。粒子孔填充和其他技术可最大程度地保护粒子的图像信息免受损失，并确保包含噪声的粒子图像可以一步完成粒子识别。同时，本文提出的方法更准确地确定了粒子图像的边缘，并为后期的流场运动计算提供了更可靠的粒子图像。对于本文提出的方法，基于Visual C ++平台自主编程，通过计算机模拟和SiO生成粒子图像。并为后期的流场运动计算提供了更可靠的粒子图像。对于本文提出的方法，基于Visual C ++平台自主编程，通过计算机模拟和SiO生成粒子图像。并为后期的流场运动计算提供了更可靠的粒子图像。对于本文提出的方法，基于Visual C ++平台自主编程，通过计算机模拟和SiO生成粒子图像。通过实际的PTV实验，分别对₂个直径为500 nm的粒子图像进行了验证和分析，获得了精度较高的识别结果。

图形摘要

粒子图像的预处理是PTV处理的前提，其水平直接影响流场的最终分析精度。根据PTV实验所捕获的粒子图像，作者提出了一种新的预处理方法，为后期计算流场运动提供了更为可靠的粒子图像。