The identification of tensile and shear cracks during rock rupture is of great significance to the rock rupture mechanism and early warning.This paper proposes a image classification and recognition algorithm based on the combination of fused features and support vector machine classifiers and a deep learning based image classification and recognition algorithm for tensor-shear crack thermal infrared images in granite for uniaxial compression experiments.Focusing on the identification and classification of infrared images of tensile and shear rupture during uniaxial loading experiments in granite,and develops the following aspects of the work:1)Data preparation: The experimental data are presented using granite infrared rupture images,and images with a single cracking feature(one image with only shear or tensor cracks)are selected to form the original dataset.The original dataset is then augmented(images are flipped or mirrored,randomly rotated)to form a database of 1000 images of clipped cracks and 1000 images of tensor cracks.2)Infrared cracked image classification based on image feature extraction algorithm:The grayscale co-occurrence matrix features,directional gradient histogram features,and local binary mode features of the two cracked IR images were extracted in the MATLAB R2020 a environment,and multiple features were combined into fused features.Single and multiple features were combined with support vector machines to classify the tensile shear cracked infrared images,respectively,and the classification results showed that the multifeature fusion yielded better classification results than the single feature.3)Infrared crack image classification based on deep learning: The deep learning pretraining networks Alex Net,Goog Le Net,and Squeeze Net are used,which are fine-tuned based on the idea of transfer learning,and then the infrared cracked images are classified.The experimental results show that the accuracy of infrared cracked image classification based on migration learning is much higher than that of traditional image classification methods.Figure 42;Table 6;Reference 51... |