Electric Equipment Image Recognition Based on Deep Learning and Random Forest

In order to analyze and recognize mass unstructured multimedia data in the electric power department automatically, we propose a new method which applies random forest to classify the electric equipment images by using features extracted by deep convolutional neural network. To be more specific, the CNN-based AlexNet model is used to extract features from the electric equipment images firstly. Then, benefitting from the achievements of digital image processing technology, pattern recognition technology, and machine learning technology, random forest is applied to classify the electrical equipment into different categories based on the deep learning features. Moreover, in order to reduce the feature redundancy, a fisher's criterion based method is proposed to select features, which are much more effective for random forest classifier than the traditional feature selection method. An electric equipment image database which contains eight thousand and five hundred electric equipment images is constructed to test the efficiency of the proposed method. There are five types of electric equipment: insulators, power transformers, breakers, power poles, and power towers in the database. Research indicates that the recognition accuracy of the proposed method is 89.6%, which is 6.8% higher than that of the softmax-based deep convolutional neural network and 12.6% higher than that of the traditional random forest. Furthermore, it can effectively eliminate the effects which are produced by the complex background. In conclusion, the proposed method can meet the actual demands of the electric power department, and it provides a new solution for intelligent analysis and recognition of unstructured mass electric equipment images. © 2017, High Voltage Engineering Editorial Department of CEPRI. All right reserved.