42CrMo single-phase metallographic structure segmentation method and 42CrMo single-phase metallographic structure segmentation system based on deep learning

Abstract

The invention discloses a 42CrMo single-phase metallographic structure segmentation method and a 42CrMo single-phase metallographic structure segmentation system based on deep learning. The 42CrMo single-phase metallographic structure segmentation method comprises the steps of dividing a 42CrMo single-phase metallographic structure image database with a true value label into a training set and a test set; then adding convolution layers in the U-Net model to six convolution layers, removing an image overlap-tile strategy in the U-Net model, adding normalization processing capable of normalizinginput of each layer of network to the U-Net model, obtaining an improved U-Net model, using the six convolution layers to increase the depth of the network, thus better extracting carbide particle features. An image overlap-tile strategy is deleted from an input image, image ghosting interference is avoided, the image is easier to obtain, carbide particle feature information is better extracted in precision, batch normalization processing is added, input of each layer of network is standardized, the convergence speed of the network is higher to a certain extent, the training speed is furtherincreased, and therefore, the research efficiency of the 42CrMo metallographic image is effectively improved.

Description

technical field [0001] The invention belongs to the field of single-phase metallographic image processing, and in particular relates to a deep learning-based 42CrMo single-phase metallographic structure segmentation method and system. Background technique [0002] Image segmentation is the first step in image analysis, the basis of computer vision, and an important part of image understanding. Image segmentation technology has an important application in segmenting carbide structure in 42CrMo single-phase metallographic structure. Common image segmentation algorithms include FCM fuzzy clustering algorithm (Chatzis S P, Varvarigou T A.A fuzzy clustering approach toward hidden Markov random field models for enhanced spatially constrained image segmentation[J].IEEE Transactions on Fuzzy Systems, 2008,16(5):1351- 1361.), Otsu Otsu algorithm (Nobuyuki Otsu.A Threshold Selection Method from Gray-Level Histograms, Systems, Man and Cybernetics, IEEE Transactions on, vol.9, no.1, pp...

AI Technical Summary

Problems solved by technology

Common image segmentation algorithms include FCM fuzzy clustering algorithm (Chatzis S P, Varvarigou T A.A fuzzy clustering approach toward hidden Markov random field models for enhanced spatially constrained image segmentation[J].IEEE Transactions on Fuzzy Systems, 2008,16(5):1351- 1361.), Otsu Otsu algorithm (Nobuyuki Otsu.A Threshold Selection Method from Gray-Level Histograms, Systems, Man and Cybernetics, IEEE Transactions on, vol.9, no.1, pp.62-66, Jan.1979.) and ACM active contour algorithm (M.Kass, A.Witkin, and D.Terzopoulos.Snakes: Active contour models, International J. Comuter Version, 1, pp.321-331, 1987.), although the FCM algorithm is used in many natural images and remote sensing In the image segmentation such as images, a good segmentation effect has been achieved, effectively and accurately segmenting the target object and the background object. However, the segmentation effect of the carbide in the 42CrMo metallographic image is not good, and the target carbide and the image background cannot be separated. The matrix tissue is effectively separated. The Otsu algorithm cannot effectively separate the matrix part from the carbide part. Most of the background matrix tissue is misclassified as carbide, and the segmentation effect is not good. The ACM model is used to extract the carbide edge of the metallographic image. Relatively The effect of FCM algorithm segmentation and Canny operator to extract carbide edges is better, but there are still some phenomena that the matrix is ​​misdivided into carbides. Because the edges of carbides are blurred, some carbides cannot be completely extracted. The study of carbides in the image has caused great trouble, which seriously restricts the research efficiency of 42CrMo metallographic images

Images