According to combine visual navigation of real-time and robustness requirement，based on the DaVinci platform, a path recognition algorithm was designed. After analyzing the texture measure in the image of wheat harvest, the path recognition algorithm of visual navigation, on account of an improved smoothness texture feature, was put forward. The original brightness image was converted into horizontal smoothness image, by the image texture feature of the horizontal smoothness. The harvest boundary was extracted from the smoothness image by using adaptive threshold segmentation, and finally visual navigation path could be determined by Hough transform. In order to validate the algorithm in the operation of the DaVinci platform， the navigation experiments of wheat harvest by combine harvester were conducted. The experimental results showed that the algorithm had the high real-time and good robustness. Its average rate of recognizing navigation line was up to 28.6 frame per second. Path recognition effect was good in a variety of environments. Algorithm operation was stable and reliable, and to continue operating time was unlimited.