In order to correct the deficiency of manual detection in the lens production process, a new method about lens-defects type recognition based on lens characteristics is put forward to make the automatic detection possible with the combination of image acquisition. After preprocessing the acquired image, the system will get the defective parameters such as area, perimeter, diameter, and so on. The types of defects are determined according to the shape characteristic, the minimum circumscribed circle and the thresholds of the different defects. The lens templates of different regions are obtained by the size of each region, the value of pixel number per millimeter and the position of the centroid. The results indicate that this method can be used to detect the main lens-defects such as points, feathers, scratches and bubbles with an accuracy of 0.03 mm, with which the detection efficiency will be greatly improved.