Rapid Defects Detection Technology For Surfaces With High Reflectivity And Transmittance

Surface defect detection is necessary in the production process.Structured light based defect detection technology has advantages of non-contact,high accuracy and speed for which it is very suitable for the defect detection in industrial environment.This thesis focuses on rapid inspection of defects for surfaces with high reflectivity and transmittance and studies the detection system construction,nonlinear error correction and dust removal.Details are listed as follows:Firstly,the principle of structured light defect detection technology is analyzed.According to the surface characteristics of high reflective surface and high transmittance surface,detection systems based on fringe reflection and transmission are designed respectively.Then,the factors affecting the defect detection results are analyzed.Secondly,a nonlinear correction algorithm based on double response curves fitting is proposed to improve the detection speed.The nonlinear Gamma response of the system introduces periodic errors.Generally,an increasing number of fringes are projected to eliminate the nonlinear errors,which will take more time.In this thesis,the phase error and modulation error introduced by nonlinear Gamma response are analyzed theoretically,and the correctness of theoretical analysis is verified by simulations.Considering the defocus effect in reflection system,an improved nonlinear response model is proposed,based on which a new nonlinear correction algorithm is proposed.Either four frames of uniform grayscale images or single frame of encoded image is required to project for correcting the nonlinear response.Compared with the existing methods,the proposed method has the highest accuracy.The root mean value of modulation errors are reduced by 29 times after compensation.Combined with this method,the result of three-step phase shifting is better than that of four-step phase shifting,which reduces the number of phase shifting fringes and improves the detection speed.Thirdly,a dust removal method based on multi-view fusion is proposed.Dust particles are too close to the digs in shape,which is difficult to be distinguished by existing methods.So dust is an important error source in defect detection.In this thesis,the structured light modulation analysis technique and polarization imaging are combined to distinguish dust and defect from each other by their different polarized scattering characteristics.The scattering models of dust particles and digs are established and the scattering matrixes of scattered light intensity from different scattering sources are obtained.After analyzing the influence that the scattering angle has on the scattered light intensity,a dust removal method based on multi-view fusion is proposed.The effectiveness of the method is verified by experiments.