基於視覺之圓形和三角形交通號誌即時偵測與辨識系統

Abstract

本文提出了一種先進駕駛輔助系統（ADAS）的視覺雷達系統的設計與驗證，檢測和分類在實際交通環境中的圓形和三角形的交通標誌。它使用從移動的車輛由相機拍攝的圖像。該算法處理兩個任務即。 （ⅰ）檢測的可能的交通標誌，並從所檢測的候選該交通標誌（ii）的分類。 該方法包括三個主要階段：第一階段是基於顏色分割，提取含有該交通標誌的視頻流的圖像，第二階段是根據形狀和特徵提取，提取含有該交通標誌的視頻蒸汽圖像通過執行霍夫變換，的Canny邊緣檢測，輪廓檢測，以及第三和最後一個階段中，我們繼續對進行分類的交通標誌作為警告，強制性和方向標誌和顯示這些分類的交通標誌給駕駛員。

This thesis proposes the design and verification of the vision radar system for Advanced Driver Assistance System (ADAS) that detects and categorization the circular and triangular traffic signs in Real Traffic environment. It uses images taken by a camera from a moving vehicle. The proposed algorithm deals with the two tasks viz. (i) detection of probable traffic signs and (ii) categorization of the traffic sign from the detected candidates. The approach consists of the three main phases: the first phase is to extract the video stream images containing the traffic signs based on the color segmentation, the second phase is to extract the video steam images containing the traffic signs based on the shape and feature extractions performed by the Hough transform, Canny-edge detection, Contour detection, and the third and the last phase, we proceed on to categorize the traffic signs as Warning, Mandatory and Direction signs and display those categorized traffic signs to the driver.