To solve the problem of mechanical parts recognition in the factory, a method of applying Support Vector Machine and combined moment to recognize mechanical parts is proposed. By extracting the Hu invariant moment of the image, combining the first and second moments into φ12, combining the third and fourth moments into φ34, combining the fifth and sixth moments into φ56, and ignoring the seventh momentφ7, the new method is constructed. The improved algorithm reduces feature dimension and recognition time, and at the same time improves recognition accuracy. The experiment comparing classification effect of the two methods shows that the recognition rate of the method extracting Hu invariant moments as features is 82.3%, while the recognition rate based on the combined moment as features is 94.1% which is higher than the former.