98%。One way of navigating autonomous vehicles to perform field operations is using a vision of crop rows detection system. A new algorithm for rows recognition was presented based on machine vision and randomized method. During image pre-processing, a vertical projection method was employed to estimate the position of crop rows. Moreover, the randomized method was used for crop rows detecting. A candidate line was determined by two different localization points, which were selected from data space randomly. The threshold rule was adopted to further determine whether the candidate line is the desired one. The test results prove that the rows detection method is suitable for different crops. The average time required for a 640×480 pixels color image is 220ms and the recognition rate reaches 98%.