Search On Texture Defect Detection Algorithm Based On Image Processing

The way of defect detection of traditional mobile phone shell is to rely on artificial eye detection,which can not meet the high standard requirements of automated production.Therefore,it is necessary to study the algorithm of intelligent automatic inspection to improve the efficiency and accuracy of defect detection.This thesis mainly takes matte surface texture background with regular mobile phone shell as the research object and then combining with the current research achievements in the field of machine vision and improving the existing method to proposed a defect detection algorithm based on image processing,the research of this algorithm is divided into two parts:texture suppression and defect segmentation.The main research contents are as follows:Firstly taking the frequency domain analysis of the texture,under the premise of studing and comparing several typical texture analysis technique,using texture spectrum method to the relevant information of texture.The analysis of the spectrum showed that there are two parts texture in image:multi period noise texture and rule line texture.It provides the direction and basis for the subsequent texture suppression algorithm.Followed by the research of Fourier frequency domain filtering to suppress multi cycle noise texture algorithm.According to the characteristics of periodic texture in the frequency domain,the algorithm using low-pass filtering in the frequency domain to suppress the periodic texture.During the experiment with the lowest frequency highlights data in frequency spectrum and calculation formula of the passband to optimize and improve selecting the cutoff frequency of the filter.The experimental results show that using Butterworth low-pass filter with the bandwidth of₀D?28?10to take the frequency domain filtering for image can retain the defect details to the full and filter out more multi period noise texture.Then study Gabor filter texture suppression improved algorithm based,inspired by the inhibition from Gabor filter theory and prove the feasibility of its texture suppression in principle.After that improved the Gabor filter from two angles of direction and scale.The improvement strategies of direction is designing a new way which combined Hof transform method with gray statistics to detect linear texture and obtain the texture direction.And the scale improvement is narrowing frequency range by analyzing the characteristics of filter parameters.The experimental results show that this algorithm behave in linear texture filtering.Finally,a defect segmentation improved algorithm based on threshold and morphology is studied.Using three kinds of typical defect segmentation algorithm to take defect detection experiments on image which was afte rtexture suppression and find that in correct rate and stability the segmentation of statistics behaving well;Besides the segmentation algorithm is improved by using morphology compensate over segmentation in the segmentation results.The structure element of morphology is designed according to direction of 2D Gabor filter which was used in the texture suppression,and verify that the improved algorithm can defect segmentation stably and accurately.