- 學位論文
利用OpenPIV進行粒子移動軌跡影像分析之初探
The Preliminary Study on the Images of Particle Moving Tracks using OpenPIV
摘要
現今土木領域中，結構物設計時不僅考慮地震力，也需考慮風力所造成的影響，透過風洞試驗可大致了解建物受風情形，當風洞實驗中產生紊流流場，其局部細節很難以透過數值模擬，但可以透過在實驗中放煙來進行粒子拍攝，以PIV粒子圖像測速法程式分析出移動量及方向，進而還原出真實流場變化情形。 研究的目的為建立粒子移動軌跡影像分析之基礎知識，在未來的水洞試驗、煙流試驗，透過添加粒子於流體中，藉由雷射等光源照射下，以高速攝影機進行拍攝粒子反射所產生光點，可得出兩張不同時間步的粒子圖片，利用PIV繪製出的速度場圖，可得出流場變化量達到流場的判讀。 研究主要利用互相關對兩張不同時間步的圖片做辨識，將取得的圖片進行影像強化及灰階等前處理，將圖片以二維矩陣代表，將圖片切割成多個一致的尺寸，對每個數列以互相關計算關聯性，得出關聯性最高的點，繪製出關聯性的三軸圖，可推斷出移動方向繪製出一個向量，將其他切割的圖片都做互相關，變求得整體的速度場圖。 研究案例分為3個部分，第一部分以Python生成不同大小的圖片，比較各個模型參數的關係，第二部分以FDS模擬建物周圍風場，第三部分以Tecplot模擬街谷效應，判斷成果上能判斷出主要流場，細部流場可透過將圖片切割成更小尺寸或將拍攝範圍縮小，進行修正。 後續除了提升模型適用性及準確性外，期望能延伸應用到其他風工程實驗的應用中，建立不同拍攝圖片的自動輔助判斷系統，透過將多個連續時間步的圖片進行辨識，比較各個速度場圖以求得較複雜的流場變化。
並列摘要
Nowadays, the influence of not only seismic force but also wind force must be considered when design of structures. Wind tunnel tests can help us roughly understand the wind effects on buildings. When turbulent is generated in wind tunnel experiments, its local details are difficult to measure by numerical simulation. We can release smoke during experiments to shoot particle photographs, and use PIV (Particle Image Velocimetry) program to analyze the moving distances and directions of particles. Then, the real flow field changes can be reestablished. The purpose of the study is to establish the basic knowledge of particle trajectory image analysis. Adding particles in future water tunnel and smoke in wind tunnel tests, we can use high-speed camera to capture the reflection of particles illuminated by laser. Two particle pictures at different time steps can be obtained, and the velocity field diagram drawn by PIV can be used to obtain the interpretation of the flow field change amount. The concept of the study is using cross-correlation to identify two pictures at different time steps. The research performs pre-processing such as image enhancement and grayscale on the obtained pictures, represents the pictures by two-dimensional matrix, and cuts the pictures into multiple consistent sizes. The cross-correlation of each series is calculated, a three-axis diagram of relevance is drawn, the point of highest correlation can be identified, and then the velocity vector can be determined. The process repeats for every picture cut to obtain the whole velocity field. The case study consists of three parts. The first part used Python to generate images of different sizes and compared the relationship between each model parameter. The second part used FDS to simulate the wind field around a building, and the third part used Tecplot to simulate the street valley effect. The main flow field can be identified. However, the detailed flow field needs to be corrected by cutting the image into smaller sizes or reducing the image capturing range. In addition to improving the applicability and accuracy of the model in the future, it is hoped the model can be applied to other wind engineering experiments. An automatic auxiliary judgment system for model parameter selection according to different image characteristics and the ability to identify velocity fields of multiple consecutive time step images are desired future research directions.