Aiming at the problems of low recognition precision caused by weak features such as micro-particle foreign objects

chromatic aberration

micro-deformation and scratches on complex surfaces

which are greatly affected by peripheral illumination and few samples

a weak feature defect recognition method based on machine vision

multi-view fast defect target source finding and fusing with intelligent pose adjustment feature enhancement is proposed.Small sample data are expanded by offline and Mosaic—6 online fusion data enhancement methods.enhanced intersection over union(EIOU)loss function and squeeze-and-excitation(SE)attention module are used to optimize the model

and multi-state data acquisition is achieved by four-light source combination projection

which improves the convergence speed of the model and enhances the feature extraction ability of defects

so as to achieve fast defect"source finding"localization.After focusing on the "source" defect position

the six-degree-of freedom manipulator and multi-angle pose adjustment enhancement are used to reveal defect characteristics and accurately identify the defect.Experiment results show that this method has high recognition precision and generalization for weak feature defects

and can meet the detection requirements of small samples and weak feature defects on complex curved surface parts.