Abstract: According to the light scattering prinliple and visual inspection method, a defect inspection system of spectacle lenses was researched and developed, which was based on the machine vision. Image acquisition, simple image processing, grading and sorting of glasses were done in the system, in order to achieve the classification of glasses. The system adopted parallel institution to run. Low-angled-ring-light-emitting-diode (LED) was used for forward illumination. Clear defect-images of speckle, impurity, feather and so on, which stayed in or on the surface of the glasses were got. Normalization algorithms were used for image processing and glasses grading, and the identifying information of various defects was acquired. The experiment results show that the detection system can detect all of the lens-s defects. The system has the detection accuracy of 0.038 mm with the average detection time of each glass less than 2 s.