本文主要研究如何從網球視訊中偵測出場線並據以重建球場影像。網球場線一般都為白色，然而白色特徵會隨著不同場地以及天候變動而改變，我們提出適應性的強度門檻值，以增加白色場線偵測的可靠度。由於Camera 拍攝時會有各種可能的變動，如平移(panning), 縮放(zooming)等，會使得部份場線消失；除外，比賽環境的外在因素，如紅土球場會因紅土覆蓋而影響場線偵測。為了解決這些問題，本文提出在Randon transform 空間下之估測隱藏線的方法，以精確的重建球場影像。This paper presents an efficient algorithm to detect court lines of tennis videos and reconstruct tennis court image accordingly. Court lines are usually white. However, the luminance feature of white lines may vary with different stadiums and weather change. In this paper, we propose an daptive thresholding scheme to improve the robustness of while line detection. In a video sequence, some court lines may disappear due to the camera motion such as panning and zooming. In addition, the court lines are often corrupted by noise such that they cannot be detected completely. To solve the so-called hidden lines, we present an effective Radon-space method to estimate the hidden lines and thus to recover the court images exactly.
