採用改良式橢圓形感興趣區域與Sobel邊緣偵測之低躁動車道線偵測
No Thumbnail Available
Date
2013
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
本論文使用垂直方向的Sobel邊緣偵測和雜訊濾波器來改善車道線追蹤時所產生的躁動問題。因為改善了躁動問題，準確率也獲得了改善。此外透過縮減影像空間處理的區域和合理的縮減橢圓形感興趣區域的大小，進而有效地提升了車道線偵測系統的處理速度。
透過實驗分析比較結果，在晴天、夜晚與雨天的情況底下皆能獲得有效的改善。實驗使用640 × 480大小的影片測試，每秒約可處理55~60張畫面，提升了約71%左右，整體的準確率方面也由原先的96.19%，提升至97.07%。
In the paper, we use the vertical Sobel edge detection and a noise filter to solve the problem of by pulse for the tracking mode of lane detection. Since the problem of by pulse is effectively solved, the lane detection accuracy is increased. Furthermore, we can effectively improve the processing speed of lane detection system, by reducing the image space and the elliptical ROI size. In experiment results, the proposed method can effectively solve the problem of by pulse in daytime, night and rain situations. The test video size is 640 × 480. The processing speed is about 55~60 frames per second. Compared with the previous method, the proposed algorithm can speed the processing of frames up to 71%, and the total accuracy is increased from 96.19% to 97.07%.
In the paper, we use the vertical Sobel edge detection and a noise filter to solve the problem of by pulse for the tracking mode of lane detection. Since the problem of by pulse is effectively solved, the lane detection accuracy is increased. Furthermore, we can effectively improve the processing speed of lane detection system, by reducing the image space and the elliptical ROI size. In experiment results, the proposed method can effectively solve the problem of by pulse in daytime, night and rain situations. The test video size is 640 × 480. The processing speed is about 55~60 frames per second. Compared with the previous method, the proposed algorithm can speed the processing of frames up to 71%, and the total accuracy is increased from 96.19% to 97.07%.
Description
Keywords
車道線偵測, 邊緣偵測, 橢圓形感興趣區域, 霍夫轉換, Lane detection, Edge detection, Elliptical ROI, Hough transform