A novel vision-based road detection method is proposed in this study to realize visual guiding navigation for Autonomous Land Vehicles (ALV). The captured image was first segmented into road region and non-road region by using Otsu thresholding algorithm. Subsequently, the extracted Canny edges located in deep position of road region and non-road region would be removed by filtering so that the recognition of road boundary could not be affected by mussy edges existed in the image. In order to improve the performance of road detection, the dynamics of ALV and the Monte Carlo Method was taken into account to associate the possible road boundary in different image frames. The method proposed in this study is robust against strong shadows, surface dilapidation and illumination variations. The real world experiment on road detection has demonstrated that the proposed method is feasible and valid.