According to various wheat field conditions in different phases, the field-edge detection algorithm based on machine vision was put forward. The camera mounted in front of the tractor could capture the images wheat field during the process of working. With the color and luminance difference of wheat field and outside the field, the position of field-edge and the candidate points of it could be detected; the slope of the field-edge was calculated by using passing a known point Hough transform. Test results showed that the method was suitable for different wheat field phases and has characteristics of fast speed, high veracity and resisting interference.