The invention discloses a pattern classification method for OPC verification. The method comprises the following steps of selecting one pattern in two comparison patterns as a first comparison pattern, and using the other pattern as a second comparison pattern; rotating the second comparison pattern and performing mirror image transformation for obtaining a plurality of equivalent comparison patterns; dividing the first comparison pattern and each equivalent comparison pattern, determining a comparison area range and selecting all related patterns which form the comparison patterns; performing rectangular segmentation on each related pattern for forming a plurality of segmented rectangles; extracting the characteristic value of each segmented rectangle; successively comparing the characteristic value of each segmented rectangle of the first comparison pattern with the characteristic value of each segmented matrix of the each equivalent comparison pattern of the second comparison pattern, and performing classification according to the comparison result. The pattern classification method can realize classification of the patterns with a rotating relation or a mirror relation and small differences to the same kind, thereby reducing the number of repetition times of the same verification result in a report, and reducing checking difficulty.