Surface defects of metal strip are important factors which affect the quality of the metal strip. How to inspect metal strip surface defects on-line is an important means for the enterprise to control and to improve the quality of metal strip products.Fully considering the consistency of the system, we designed the on-line metal strip surface defects inspection system which includes image acquisition, real-time judge, image pre-processing, defect segmentation, feature extraction and classification. Simulated the production scene in a surface defect detection experiments, we got several achievements: taking into account the characteristics of metal strip surface defects, in laboratory conditions, designed the hardware systems of metal strip surface defects inspection system, effective simulated the production process, laid a foundation for the verification of system in production environment; studied real-time inspection method, used semi-closed image collection methods to overcome the impact on gray image brought by environment, the noise of the environment light has been controlled effectively; designed an improved smoothing filter, experimental results showed that the algorithm eliminated the noises in the images very well; for the acquisition of six typical surface defects, designed a classification algorithm based on BP neural network. Overall recognition rate reached 88.7 percent.Through this research and exploration, provide the necessary technical support for the inspection of metal strip surface defects. |