Abstract: Defective glass components will explode frequently after being put into operation, which threatens the stable operation of power grid. It is necessary to remove the defective glass as much as possible on the production line. However, the existing defect detection methods of glass is not effective enough as expected. Thus, a glass defect inspection method based on machine vision is proposed. After analyzing the principles of detecting glass defects with uniform light sources and laser point sources respectively, the proposed method uses both light sources in combination along with the machine vision inspection method. We also developed an image recognition algorithm and established a set of indexes of defect characteristics, and eventually realized automatic identification of bubbles and crack defects in the plates and ridges of glass insulators. Experiments show that the proposed inspection method has the advantages of using uniform light sources, namely the simple layout and fast acquisition, as well as the advantages of using laser light sources, namely the obvious phenomenon, simple algorithm and high accuracy. The results indicate the potential of practically applying the proposed method.