Research on Transmission Line Icing Image Classification Method Based on AlexNet and SVM

This study proposes a method for classifying transmission line icing images by combining AlexNet, a deep convolutional neural network, with a Support Vector Machine (SVM) classifier. By leveraging AlexNet's feature extraction capabilities and combining them with manually extracted Local Binary Pattern (LBP) features, the model achieves a classification accuracy of 96%. Modifications to AlexNet's architecture improve generalization and reduce overfitting. Experimental results demonstrate the model's robust performance, offering a promising approach for accurate icing classification and enhancing power grid safety.