The problem of detect navigation line by machine vision under non structural environment and complex background in greenhouse was addressed. After analyzing the distribution characteristics of the road images in tomato ridges based on HSI color space, a method was proposed for extracting the sensitive region of the heat pipe. Firstly, the binary images were generated from the road images by using Otsu segmentation method based on I histogram. The discrete set of navigation points was obtained by using edge extraction algorithm in accordance with the binary images. And then, the two sidelines of the heat pipe were obtained by using the least square method to fit the discrete points, on the basis of this, the navigation line detection algorithm was proposed. Finally, the influences of the illumination in homogeneity and the different covering on navigation line detection were investigated. The experimental results indicated that in comparison with Hough transform, the algorithm was simple and fast, as well as robust with illumination in homogeneity, the edge information of target region could be extracted accurately and the recognition correct rate under different covering was 91.67%. 