A Multi-structure Elements Based Lane Recognition Algorithm

摘要

W artykule opisano metodę rozpoznawania linii na jezdni z poruszającego się pojazdu. Metoda bazuje na opracowaniu modelu morfologicznego multi-strukturalnego linii na podstawie danego obrazu oraz zastosowaniu transformaty Hough w celu wyznaczenia parametrów. Uzyskano w ten sposób zwiększenie szybkości w przypadku złożonego otoczenia, przy jednoczesnej niskim stopniu skomplikowania.%The existing traffic lane recognition algorithms have the weaknesses of low recognition ratio, bad robustness and real-time, for overcoming these drawbacks, this paper proposed an algorithm of lane recognition based on multi-structure elements model of morphological. In the algorithm, the region of interest (ROI) is extracted from the original image, which is detected by the operator of Canny. After that, the lanes are extracted by the structure elements, which have similar characteristics to that of lane model. Several lines are detected by Hough transformation, and choose the parameters to reconstruct the traffic lane. The experiment results show that this algorithm is simple, has better robustness, and at the same time, can efficiently detect the lane mask accurately and quickly.