Detection of geometric shape for traffic lane and mark

Traffic lane and mark are geometric shapes of road. Detection of geometric shapes is an essential component of autonomous urban driving system. In this paper, we introduce a method to detect geometric shape for traffic lane and mark. This method contains four steps. First, we apply HSV color space, histogram equalization and Otsu algorithm for image preprocessing. Then we use Canny algorithm and Progressive Probabilistic Hough Transform algorithm finding an optimal position of lane in image. Then by using Kalman filter model, we update and track lane marking lines. Finally we use template matching method to detect shape of traffic mark. We introduce normalized cross-correlation to recognize the traffic signs. Experiment results demonstrate that the proposed scheme can detect geometric shape for traffic lane and mark effectively.