Electric Power Equipment Image Recognition Based on Deep Forest Learning Model with Few Samples

Abstract

Power system equipment have many characteristics, such as many kinds, single color, similar appearance and complex environment of transmission and substation. These characteristics make it difficult for traditional image recognition technology to achieve good recognition results. If there are few training data samples, the performance of in-depth learning recognition declines sharply. To solve these problems, this paper designs an optimization and training recognition method based on the combination of generating confrontation network and deep forest. Firstly, the generated countermeasure network is used to expand the samples of power equipment, and then the active learning method is used to optimize the generated samples. Secondly, the traditional model-enhanced sample expansion method is used to expand the optimized samples twice. Finally, the robust network model is obtained by using the deep forest method to accurately identify the power equipment. Experiments show that this method has high recognition accuracy and is superior to many other algorithms.