To prevent an intelligent vehicle from departing the lane in the vision-based navigation, an integrated method based on image processing is proposed to detect the road boundary and lane marking synchronously in structural road environment. The feature of the road boundary is extracted by means of gradient magnitude and gradient direction of pixels. And the lane marking feature is extracted by self-adaptive threshold segmenting with region connectivity analyzing. The characteristic points of both the road boundary and lane marking are matched to the straight or crooked road models by least-squares fit. With the circular calling of detecting and tracking blocks for mass image sequences, the whole process shows a real time and high antinoise capability. All the algorithms in the paper have been tested by the videos captured from real road scenes, and the experimental results proved that the detecting method is efficient, stable and accurate.