Metal fracture image recognition method based on deep learning

Abstract

The invention discloses a metal fracture image recognition method based on deep learning. The method comprises the steps of 1 randomly dividing metal fracture image data into a training set, a test set and a verification set according to a proportion of 70%, 20% and 10%; 2 performing data expansion on the training set by adopting a data enhancement technology; and 3 initializing parameters of a convolutional layer and a pooling layer in the convolutional neural network VGGNet16 by adopting a transfer learning technology, and training and optimizing the convolutional neural network VGGNet16 onthe basis until the VGGNet16 network converges. According to the method, automatic feature extraction can be realized, and the steps of artificial feature extraction and selection in a traditional machine learning algorithm are omitted; meanwhile, in the model training process, the automatic feature extraction mode of the model can enable the model to automatically extract the features of the object to be recognized more easily, and therefore the best image recognition precision is obtained.

Description

technical field [0001] The invention relates to the field of metal fracture image recognition, in particular to a method for metal fracture image recognition based on deep learning. Background technique [0002] Metal fracture images are mainly divided into four categories: cleavage, dimple, fatigue, and intergranular. At present, there are many methods for fracture image recognition: the first one is to use gray-level co-occurrence matrix, which is a feature extraction algorithm based on statistics. The method and correlation analysis for feature dimensionality reduction have achieved good recognition results; the second is to use the defined fuzzy gray level co-occurrence matrix for feature extraction, and the classifier chooses the hidden Markov model to achieve a higher recognition rate; The third is to use Grouplet transformation to process metal fracture images, combining Relevance Vector Machine (RVM) and Kernel Principal Component Analysis (KPCA), respectively, to p...

AI Technical Summary

Problems solved by technology

Although traditional machine learning algorithms can also recognize metal fracture images well, these algorithms cannot achieve end-to-end learning in the process of training and testing, that is, from input to output, the data is input into the model, and the model automatically Extract the features in the data and generate prediction results; at the same time, manually select the features in the traditional machine learning algorithm, the quality of feature selection will directly affect the accuracy of image recognition, which may lead to the final image recognition results may not be the best the result of

Images