Research On Foreground Region Extraction And Phase Unwrapping Of Gear Interferogram Image

Since the development of involute helical gear for a hundred years,because of its advantages of stable load,high bearing capacity and low noise,it has an indispensable position in gear manufacturing.In the manufacturing process,the micrometer shape error of gear will greatly affect the noise,life and performance of gear and its system.Compared with the traditional contact measurement method,laser interferometry has the advantages of high precision,fast speed and non-contact,so it is the main development direction of gear shape error measurement.In the laser interferometry of gear morphology,the accuracy of foreground region extraction and phase unwrapping of interference image is the key problem to ensure the final measurement accuracy.Therefore,this study conducted in-depth analysis and research on several key problems,the specific content is as follows:At present,the threshold setting of common gear interference image foreground extraction algorithms completely depends on the experience of researchers,and the accuracy cannot be quantitatively reflected by rules,In this paper,an adaptive threshold foreground region extraction algorithm based on the gray level of tooth surface object is proposed.Firstly,the morphological characteristics of gear tooth surface and the difference of edge vertices were analyzed,and the region of the image was divided.Then,according to the change law of edge gray level,qualified pixels are screened through the neighborhood window and mask results are obtained to achieve the extraction of foreground region.Finally,the segmentation results of the algorithm are compared with those of traditional methods.The algorithm realizes threshold adaptive and improves the accuracy of foreground region extraction significantly.At present,Common foreground extraction algorithms are highly dependent on the image of gear eating surface,which leads to the increase of the measurement link and the low efficiency of the algorithm.In this paper,an algorithm for extracting the foreground region of gear interference image is proposed based on fringe sinusoidal characteristics.According to the sinusoidal variation characteristics of the gear interference fringe,the threshold adaptive graylevel mask was established,and the evaluation function was established to analyze the interference image fluctuation state to obtain the repair mask.By calculating the combination of two masks,the foreground region was directly extracted.Finally,the qualitative comparison and quantitative analysis are carried out by referring to the results,the traditional algorithm and the algorithm in this paper.The algorithm realizes the extraction of foreground region directly through gear interference image,and the operation efficiency is greatly improved on the premise of ensuring the accuracy of foreground region extraction.The traditional phase unwrapping algorithm is unable to unwrap some low quality gear phase area.In this paper,Res U-net convolutional neural network is used to realize phase unwrapping of gear interference images.Firstly,the interference fringe wrapping and unwrapping image data set was constructed by Zernike fitting simulation,and various noises were added for model training to verify the feasibility of the network structure.Then,the local feature extraction method of gear interference fringe wrapping and unwrapping image is used to build data set and train the model.Finally,the Res U-net network and the traditional algorithm are compared respectively with the true value.The problem that the traditional algorithm cannot unwrap low quality gear phase area is solved successfully,and the unwrapping accuracy is improved to a certain extent.