Blurred Image Restoration In On-line Measurement Of Rail 3D Morphology

In rail traffic,rails play an irreplaceable role.The rails are inevitably subject to wear during the operation of supporting and guiding the vehicle.In order to ensure the safe operation of rail transit,it is necessary to detect the wear of the rail.The traditional two-dimensional measurement method only detects based on the feature information of the image,and the two-dimensional measurement can only obtain two-dimensional information.The three-dimensional information obtained by the three-dimensional measurement cannot be realized by the two-dimensional measurement.In the online Phase Measurement Profilometry(PMP),if the measured object moves at a relatively high speed,the fringe image obtained by the acquisition system is usually a motion blurred image.The resulting motion blurred image will increase the 3D reconstruction error.In severe cases,3D reconstruction will not be possible.To apply the online PMP to the on-line measurement of the three-dimensional shape of the rail,in order to realize the clarify of the motion blur deformation fringe of the rail surface,Wiener filter algorithm,point spread function algorithm,blind deconvolution algorithm and the Richardson-Lucy algorithm were used for restoring the motion blur fringe image.Image quality evaluation was performed on the restored images by the four image restoration methods by using the peak signal to noise ratio.The relationship between vehicle running speed and restored image quality was researched,and the relationship between image restoration quality and vehicle running speed was obtained.Finally,error analysis was performed.The theoretical and experimental results show that the quality of the reconstructed image by using the Richardson-Lucy algorithm is better when realize the clarify of the motion blur deformation fringe of the rail surface.The relationship between the restoration quality of the image and the running speed of the vehicle is multiple-times.In order to improve the accuracy of PMP and prevent errors in the process of 3D reconstruction,this thesis improves the traditional Stoilov algorithm and effectively removes four kinds of singular points.At the same time,this thesis proposes an improvedregularization method to restore motion blurred fringe images.The calculation and estimation of the point spread function have been researched in many aspects,focusing on the two aspects of the point spread function.For the point spread function of successive frames of motion blurred images,it can be approximated to achieve a large degree of saving operation time and obtain a highly accurate point spread function.Using the Radon transform algorithm on the calculation of the point spread function,the obtained point spread function will greatly shorten the time and improve the accuracy in the process of regularized edge reconstruction.Accurate online 3D measurement of the 3D rail shape can be achieved by using the improved regularized image restoration algorithm and improved PMP.