Based on the corn rhizome image information, a corn field navigation line extraction method combining edge detection and area localization was proposed. Firstly, the 2G-R-B grayscale image was segmented and the binary image was obtained by using the maximum betweenclass variance. The morphological processing was combined with position/area denoising methods to improve the quality of the binary image and reduce the noise. The images were accumulated in columns to obtain the column pixel accumulation curve. The traditional method needed to set the distance threshold when extracted the feature points. Gaussian filter was used to smooth the accumulation curve and extreme value method was used to reduce the interference of pseudo feature points in maize roots and stems. When extracted the straight lines of corn stalk edges, a twosided edge discrimination method was proposed based on the image width of the furthest stalk, and the pseudoedge straight lines were effectively eliminated by scanning the closed quadrilateral neighborhood of each edge line. Finally, based on the straight line of the edge, the local area of the corn rhizome was relocalized and the false feature points were eliminated. The leastsquares linear fitting method was used to accurately extract the navigation lines. The experimental results showed that the algorithm took about 236ms to process a 1280pixels×720pixels image, and the accuracy of feature point fitting was 92%. Compared with the traditional methods, the algorithm had the characteristics of high accuracy and good realtime performance. The algorithm was still more robust in the case of lack of seedlings, more weeds, and nonstandard plant spacing. It can provide visual navigation for intelligent agricultural machinery to control corn diseases and insect pests.