A novel method was presented to detect road boundary in infrared image for Autonomous Land Vehicle (ALV) navigation in this paper. The characteristics of road infrared image were discussed in detail and the shortcomings of normal image processing methods used in road CCD image were pointed out. Because of the gradual change of temperature, the consecutive threshold technique was adopted to obtain segmentation results and calculate the gradient intensity. A new strategy combined edge and gray scale information was proposed to extracte road boundary edges and candidate boundaries. This strategy has been applied to ALV navigation and poven practical.