Research Of ESPI Fringe Orientation Extraction And Image Inpainting Based On Deep Learning

Electronic Speckle Pattern Interferometry(ESPI)measurement technology,as an important optical non-destructive testing technology,has been widely applied in many practical engineering fields.In electronic speckle interferometry,the physical quantity to be measured is directly related to the phase information encoded in the fringe.Accurate phase information extraction from fringe image is the key to the successful application of electronic speckle interferometry.Therefore,this paper carried out the following research on the accurate extraction of phase information of electronic speckle interference fringe image.The main contribution of this paper can be summarized into three aspects as follows:(1)In this paper,a method for calculating fringe orientation of ESPI image based on Convolution Neural Network is proposed.Orientation information is an important feature of ESPI fringe image,and it is useful for filtering and skeleton extraction.In order to obtain fringe direction efficiently and accurately,a fringe orientation extraction method based on Convolution Neural Network is proposed.In this method,the simulated fringe image and its corresponding orientation value are used as the training set to train the network,and the simulated fringe pattern is used as the verification set to verify the performance of the network.After the network training is completed,the ESPI fringe image can be input into the network,and its orientation information can be obtained efficiently and accurately.The advantage of this method is that the fringe orientation information can be extracted accurately and efficiently without complex parameter adjustment.(2)In this paper,a method to repair the broken ESPI fringe image based on Generative Adversarial Networks is proposed.Due to the reasons of experimental equipment and environment,the obtained ESPI fringe image sometimes appears break,which seriously affects the accuracy of extracting information.According to the above problems,a method to repair the broken ESPI fringe based on Generative Adversarial Networks is proposed.In this method,the broken image and the binary mask corresponding to the broken region are used as the training set to train the network,and the simulation broken image is used as the verification set to verify the performance of the model.After the network training is completed,the broken image and the binary mask corresponding to the broken area are input into the network,then the repaired image can be obtained.The advantage of this method is that it can repair the broken ESPI fringe quickly and effectively.(3)In this paper,the orientation calculation,filtering and skeleton extraction of the broken ESPI fringe and the corresponding repaired fringe were performed respectively.After the image is repaired,the repaired image has more accurate calculation orientation,better filtering effect and more accurate skeleton.