An image acquisition system is established to realize the automatic detection of resin lenses. The parameters of the system and algorithms for image pre-processing, feature extraction, defect recognition, etc. are investigated. The image acquisition system is built according to the size of lens and detection precision of defect. Image mosaic is realized by camera calibration. The defect images are acquired through image pre-processing. The algorithms of defect feature extraction and classification are finally discussed. The experimental results indicate that the detection precision of the system is 0.02 mm and the recognition rate reaches 96.5%, which can meet the system requirements for lens detection.