以影像為基礎之交通標誌標線自動化辨識與幾何重構

交通標誌與標線為公路養護工程中重要的項目之一，但目前在巡查作業上多仰賴人工進行，導致公路巡查效率不彰且資料更新緩慢。本研究發展以影像為基礎之標誌標線自動化辨識及測繪技術，以複合式門檻建立適應戶外環境之候選物偵測方法，接著以具尺度、平移及旋轉不變性之特徵進行形狀分類再透過支持向量機辨識標誌圖案內容，完成辨識後以時間連續性與成像幾何為依據將針對相同標誌標線之連續辨識影像分段，最後以空間前方交會解算標誌標線之物空間坐標。由實驗結果顯示：在辨識階段中標誌、標線之形狀分類正確率以獨立個數計皆可達100%；以影像張數計則可個別達98.2 %與99.7 %，標誌之圖案辨識正確率以獨立個數計可達95.0 %；以影像張數計可達89.4 %，並且於日、夜間皆能獲得穩定的成果。而藉由物件追蹤能改正分類錯誤之物件類別並提供基線長足夠之共軛像對以進行空間坐標解算。

關鍵字

標誌標線辨識；物件追蹤；空間前方交會；支持向量機； DtB向量

並列摘要

The maintenance of road signs and markings is an important issue for road user safety. An automatic recognition technique can provide road information in short time, making maintenance tasks more efficient. This study developed an image-based approach for automatic traffic signs and markings recognition and reconstruction. The proposed algorithm is translation-, scale-, and rotation variant, and is thus capable of detecting and recognizing traffic signs and markings in an outdoor environment. The object tracking technique can be used to segment image sequence adaptively and to provide conjugate photos for space intersection. Based on the results from a field experiment, it has been demonstrated that the successful detection rate reached 100 % for traffic signs and 80.9 % for markings. The classification accuracy of shape reached 100 % for both traffic signs and markings, and the recognition accuracy of traffic sign pattern reached 95.0 %. Consequently, an image-based automatic recognition and reconstruction of traffic signs and markings is achieved by algorithm proposed in this study.

並列關鍵字

Sign recognition ； Marking recognition ； Space intersection ； DtB method

參考文獻

Barnes, N., and Loy, G., 2006. Real-Time Regular Polygonal Sign Detection (Vol. 25). Springer Berlin Heidelberg.6.

Barnes, N., and Zelinsky, A., 2004. Real-time radial symmetry for speed sign detection. In Intelligent Vehicles Symposium, 2004 IEEE. pp. 566-571.

Benallal, M., and Meunier, J., 2003. Real-time color segmentation of road signs. In Electrical and Computer Engineering, 2003. Canadian Conference on vol.1823. pp. 1823-1826

Cortes, C., and Vapnik, V., 1995. Support-vector networks. Machine learning, 20(3). pp. 273-297.

Duda, R. O., and Hart, P. E., 1972. Use of the Hough transformation to detect lines and curves in pictures. Commun. ACM, 15(1). pp. 11-15

國際替代計量

以影像為基礎之交通標誌標線自動化辨識與幾何重構

全文下載

主題瀏覽