Dirt detection on camera module using stripe-wise background modeling

With the rapid development of smart phones, tablet computers and other digital products, the demand for camera modules is growing strong. However, in the production line, it is difficult to ensure the workshop is completely clean. Lens of the camera module is easily contaminated by dust or oil matter, which will bring serious influence on imaging quality. Therefore, we need to detect automatically the presence of dirt and blemish on lens, in order to improve the overall quality of camera modules. Considering expensive manpower cost, artificial detection is not economical and detection accuracy can't be guaranteed. Through analysis of the imaging mechanism and the reasons of dark corner formation, we propose a novel and efficient method for automatic dirt detection based on stripe-wise background surface modeling. Experimental results show high detection accuracy and robustness of the proposed approach, even in case of significant dark corner and high level dirt.