This thesis presents the road lane detection algorithm. Our proposed algorithm is based on a probabilistic approach," Bayes Decision Rule" that consists of a prior probability model and a likelihood function. This approach is applied to each divided block in the image region and can extract the prominent blocks by using the prior knowledge of road lane direction. This known knowledge can make a probability model and compose the effective posteriori probability function with the gradient information of image. We use this posteriori function to detect the road lane candidate blocks at a coarse level.
Among these blocks we must select the road lane. We propose the block snakes that has a block direction energy and a connecting line direction energy as internal energy and a gradient magnitude energy as external energy. This block snakes can decrease the noise effect and computation time compared to the point snakes based on pixel. We also propose the contour following algorithm that can find the multiple contours by only one time search of the image.
By the prior model updating using the detected blocks direction information, we can estimate the front road direction and increase the quality of detection procedure.
