To overcome shortages of the existing algorithms for crop rows detection, such as the noise interference and slow processing speed, an algorithm based on constraint of liner correlation coefficient was proposed for crop rows line detection. Firstly, according to the characters of farm land images, ROI area was processed to increase the speed of image processing. Secondly, crop row numbers and edge information were detected by vertical projection method. Finally, an algorithm based on constraint of liner correlation coefficient was adopted to fit the feature points of crop to realize line detection. In order to fit the feature points, the N points were removed, which were the farthest from each side of fit line, until the fitting line met the requirement of liner correlation coefficient r. By comparing with Hough transform and random detection algorithm, the results showed that the proposed algorithm had the advantages of fast detection and strong anti interference. The results of navigation tracking experiment showed that this navigation system could track the target path accurately and effectively. The maximum lateral deviation was no more than 4cm and 7cm at the speed of 0.6m/s and 1.2m/s respectively.