Defect Detection In Periodic Texture Surface

Machine vision technology has been widely used in the quality inspection of the surface of industrial products.The defect detection algorithm is the core of machine vision technology.There are many kinds of products with periodic texture.Therefore,it is of great significance to study defect detection algorithms on periodic texture surfaces.Based on the mechanism of linear superposition between periodic texture and defects,this paper presents a study of how to remove the periodic texture and retain the defect information in the image reconstruction.The application of image reconstruction method,two-dimensional Fourier transform and singular value decomposition,in periodic texture surface defect detection is explored in depth.For the image reconstruction method based on two-dimensional discrete Fourier transform,the high energy frequency component in the amplitude spectrum of the texture image corresponds to the periodic texture,therefore the high energy frequency components can be removed by the filter.The gray level of the periodic texture in the reconstructed image is suppressed well,meanwhile the defect is enhanced relatively.Finally,it is easy to split the defect from the periodic texture.Furthermore,by normalizing the amplitude spectrum,the process of constructing the filter is removed,while the effect of detection is better.In the image reconstruction method based on singular value decomposition,almost the same periodic texture image can be obtained by selecting the appropriate number of singular values to reconstruct.On this basis,the periodic texture information in the original image can be removed by self-subtract,and then the defect information can be preserved.The experimental results show that these two algorithms have good effect on the defect detection of cigarette packaging with periodic texture characteristics.