On Determining the Image Correspondence Using Fourier Descriptors

使用傅立葉描述子決定影像間之相關

簡新府(Hsin Fu Chien)

《中華科技大學學報》 69期 (2017/01) Pp. 1-12

摘要

Invariant nature of Fourier descriptors has been addressed heavily regarding its capabilities in translational, scalable, and rotational independence of targets. This study aims to bridge the gap between its theoretical property and practical use by illustrating with professional sports application of U.S. Major League Baseball (MLB). Boundary points were obtained from two sets of pitching delivery images to get the Fourier coefficient information in determining whether or not the correspondences exist. The baseball pitcher mechanical adjustment application analyses the professional sport player’s gesture changes in order to make diagnosis and suggest for improvement. This paper utilizes the invariant property using Fourier descriptor between two captured images to determine the existence of correspondence. The purpose for conducting the experiment in this approach is to align images and aim to provide an automated coaching in gesture adjustment for professional athletes after analyzing the corresponding images taken at different time. Experimental results demonstrate the applicability of Fourier descriptors to the application domain for finding the invariance between correspondent images.

關鍵字

Invariance ； Fourier descriptor ； pitching mechanics

並列摘要

傅立葉描述子之不變性在對目標物的位移、縮放、與旋轉獨立性的能力受到大量的討論。本次研究目標在於藉由使用美國大聯盟棒球的實例橋接理論性質及實際應用，透過取得兩組投球影像的輪廓邊緣點獲取傅立葉係數並以此決定相關是否存在。棒球投手機械性調整的應用乃分析職業運動選手的姿勢改變以做出問題診斷及改善的建議，本論文使用兩組擷取影像之傅立葉描述子的不變性性質判斷其間相關的存在，使用本手段進行實驗的目的在於影像之校準並希望提供分析不同時間取得之職業選手之影像進行姿勢調整之自動專業教練建議。實驗結果證明傅立葉描述子在相關影像間之應用尋找不變性的能力。

並列關鍵字

不變性 ； 傅立葉描述子 ； 投球物理