23ms。The influence of the various illumination, weed, and shadow environment to the path recognition was investigated. A new research method for vision navigation system of agricultural vehicle in the application environment was proposed. The various navigation environments were classified automatically by neural network. According to classification results, different path recognition methods were applied in navigation system. Experimental results of the recognition and the classification prove that the method is effective and can improve the practicability and reliability of vision navigation system. The classification correct rate is 95%, and the average cost time is 23ms. 