This paper describes a new machine vision based method for detecting field crop-rows in corn. The method can be applied to guidance system in agricultural robots or autonomous vehicles. Instead of using the classical Hough transform, a method based on curve fitting and iteration was developed. This method was proved to be more robust with respect to noise caused by the large leaves and can provide better accuracy. In addition, it features reduced computational complexity of the image-processing algorithm. The algorithm operates on binary images, which were segmented from RGB images based on color information. The segmentation was adapted to various lighting conditions. The interference caused by weeds and crop residuals was decreased by using a spatial filter. It is also robust with respect to special conditions, for example, the occurrence of the end of the rows or weed presence. A fifth order low pass filter was used to increase the efficiency and accuracy of the detecting algorithm. The algorithm was tested on 100 images, among which 98 percent received acceptable results. A traditional Hough Transform based row detection method was chosen as a reference, the method of this paper was proven to have a better result in both accuracy and time costs. The processing time was 0.13 second per image on average.