A machine vision system was constructed using industrial cameras, lenses, computers, and specialized image processing software according to industrial inspection speed and accuracy requirements of groove width detection of right-angle trapezoid groove of cylindrical part. Groove width detection method of right-angle trapezoidal groove of cylindrical parts was design based on image processing algorithms. Comparative experiments showed that the detection accuracy of the system in groove width of cylindrical part can achieve micron order, which is better than manual measurement using a dedicated groove width measurement tools; the measurement results are not influenced by subjective factors; the efficiency is very high.