Reserach On The Nondestructive Testing Technology Of Surface Defect Of Mirror Object Based On Fringe Reflection Method

With the rapid development of manufacturing industry and precision instruments constantly updated and upgraded,the requirements for the quality and performance of specular reflective materials which are widely used in aerospace,automobile manufacturing,household appliances and other fields are becoming more and more stringent.At present,among the detection methods for surface of mirror objects,the surface measurement technique based on fringe reflection has attracted great attention of researchers because it has the advantages of high sensitivity,fast detection speed and low cost.As with other detection methods,improving the measurement accuracy is always the focus of fringe reflection technology.In this paper,the phase unwrapping technique and three-dimensional reconstruction technology are studied,and the sparse representation method is innovatively proposed for fringe reflection detection.(1)In practical detection,LCD are used to display encoded standard stripe images,CCD camera captures distorted fringe images reflected from the surface under detection,and the phase relating to the surface under detection can be obtained by the phase demodulation technology.In this paper,the mathematical relationship between phase and gradient is derived theoretically,the geometric calibration principle and process of the measuring system are introduced,which lays the foundation for three-dimensional shape reconstruction.(2)To solve the problem with phase unwrapping technique,sparse representation is proposed for phase unwrapping.The mathematical model is established,and the unwrapping phase is solved by gradient descent algorithm,the results of numerical simulation and comparison with the traditional phase unwrapping algorithm show that,the precision of unwrapping phase obtained by the sparse phase unwrapping algorithm is higher,there are not this phenomena such as large error and pull wire in the phase distribution map,which indicates that sparse representation phase unwrapping can suppress noise effectively.(3)There may be non-integrability and various image noises in gradient domain when three-dimensional surface is reconstructed by gradient integral,which will lead to increase of the reconstruction error and misjudgment of defects.In view of this problem,sparse representation is introduced into the curl calculation equation in this paper,and the curl is expressed as a linear combination of the given redundant curl dictionary and the sparsest solution of the gradient error,which overcomes the non-integral condition and the interference of various noises.In order to verify the validity of the algorithm,numerical simulation is carried out separately under the conditions of Gaussian noise,Rayleigh noise and uniform noise.The results demonstrate that the sparse representation gradient integral can effectively suppress various types of noises and markedly improve the accuracy of scratch judgment comparing with the standard least square integral.Preliminary experimental results also further verify the effectiveness of the algorithm.