A detection and control methold based on the combination of the positioning and measurement in the realtime embedded machine vision systems is presented. To solve the problem that the industrial detection is of lower accuracy and can not be completed in a realtime way by embedded machine vision systems due to large computation cost, a detection method is proposed based on the point set, the Hough transformation and the leastsquare fitting algorithm. The industrial detection experimental results show the workpiece detection accuracy based on the method is better,the tolerance is higher, and the detection time is lower than 2.0ms/time.