Mobile device appearance detection method and system based on deep learning

Abstract

The application discloses a kind of mobile device appearance detection method and system based on deep learning, it is related to machine vision detection technical field.The method comprises the following steps: S1, the multi-angle appearance original image set of mobile device to be measured is acquired, and the multi-angle appearance original image set is executed standardization preprocessing;S2, the standard eigenfeature benchmark library corresponding to the mobile device to be measured is acquired;S3, by the eigenfeature decoupling network of pre-training completion, the feature decoupling processing is carried out to the standardization measured image set, and the measured topography eigenfeature map, measured texture eigenfeature map, measured illumination reflection eigenfeature map corresponding to the standardization measured image set are output;S4, based on the multi-angle space conversion matrix, all the measured topography eigenfeature map is mapped to the unified world coordinate system.The application realizes the high-precision detection of mobile device appearance defect, effectively filters out false defect interference, gets rid of the dependence on defect annotation sample, and adapts to the automatic detection demand of production line.

Images