ABSTRACT
This paper presents a novel optimisation algorithm for lane detection and fitting. We use the Artificial Fish Swarm Algorithm (AFSA) which is based on a parabolic lane boundary model to solve our boundary detection problem. Initially, the RGB road image is transformed into intensity images. We then use the Finite Impulse Responses (FIR) filter to eliminate noise. Next, the Otsu’s method is used to convert the intensity images into binary images. Thirdly, an objective function is constituted with gradient character and grey level of the binary images, coming along with the road boundaries are fitted by parabola model. The major point of the AFSA is to optimise the parameters of the quadratic parabola based on the objective function, and setting a frame’s parameters’ values as the initial value of the next frame, which is equivalent to the tracking method. Experimental results of the real-time image sequences show that the method presented in this paper is capable of robustly and accurately detecting the road boundaries on highways roads. The accuracy of the algorithm has reached a high level. The mean processing speed of each image is 37 ms.
