Abstract: In order to develop an autonomous navigation system suitable for the whole growth cycle of sugarcane in hilly and mountainous terrain, two-dimensional LIDAR was used as a navigation sensor, and a correction algorithm was designed to correct the influence of terrain on the lateral data of the radar. Due to the large differences in the shape of sugarcane in different growth periods, the traditional clustering algorithm had fixed clustering parameters, resulting in poor actual clustering effect. An adaptive clustering algorithm based on point cloud number threshold resolution was proposed. The confidence interval was introduced by using the characteristics of filtering and sugarcane linear planting. After locating the sugarcane, the improved least squares method based on slope determination was used to realize the straight line fitting of a small amount of data. Field experiments confirmed that this method could eliminate the influence of terrain and sugarcane leaves on the radar point cloud data, and the positioning error could be controlled within 10 cm. In this regard, the fitting of the walking path in the whole growth cycle of sugarcane is realized.