Research On Image Recognition Of Power Equipment Based On Improved Regional Convolution Neural Network

Patrol monitoring of power equipment is one of the important measures to protect the safe and stable operation of the power grid,and is an important work to ensure the safe and stable operation of the entire power grid.At present,the manual periodic inspection method not only consumes a lot of manpower and material resources,but also is prone to missed detection and false detection of power equipment.Although the traditional power equipment image recognition method is simple and targeted,it still requires manual feature extraction,which has no good mobility and generalization,and cannot be directly applied to real-time detection of intelligent inspection equipment.Although the Regional Convolution Neural Network has shown excellent performance in image localization and recognition,whether they can be applied to image recognition tasks of electrical equipment in small-scale image databases under complex backgrounds still needs to be verified.In order to solve the above problems,we introduce the long-short memory network structure of recurrent neural network into a faster regional convolution neural network model,and design a power equipment image recognition network model based on Faster LSTM-CNN.The long-term dependence information in the picture data of the power equipment is learned through the LSTM module to extract this special picture feature.The LSTM module can also deepen the understanding of the positional relationship of power equipment in the image and learn the structural features of the power equipment in the image.It is suitable for end-to-end detection of power equipment images and has good generalization capabilities.Finally,this paper compares the recognition efficiency of different pre-training networks used for migration learning in Faster LSTM-CNN network model.It is verified that Faster LSTM-CNN network model is superior to other R-CNN network models in the recognition accuracy of the five power equipment described in this paper,especially in the recognition of small-sized power equipment such as shockproof hammer.It provides technical support and reliable guarantee for image recognition technology based on depth learning to automatically detect power equipment in intelligent patrol inspection system.