In order to effectively classify mechanical components in embedded system, a novel method was proposed to extract edge features of mechanical component images since mechanical components have comparatively obvious edge features. First, the Kirsch operator was used to extract the edge of mechanical component binary image. Then the edge image was divided into several areas with the centroid as the center, and the modified normalized central moments of each area were calculated. The vectors including these central moments were constructed to be the edge features of mechanical components. In the experiments, the means clustering algorithm was used to classify the components based on the extracted edge features, and the effectiveness was proved by the results.
