Abstract: To improve the accuracy of microvision inspection, an edge detection method based on processing matching principle is proposed: extracting the transition region of parts processed by different machining methods; the polynomial regression mathematics models of transition region are built up and the accurate position obtained from the mathematical model. Experimental results of different samples indicate that this algorithm considered the effects by processing characteristics, and the edge recognition result can reach to the sub-pixel level by establishing the mathematical model of transitional region. The algorithm can also offer a continuous edge, which can meet the precision measurement requirements.