Power equipment recognition method and system and storage medium

Abstract

The invention discloses a power equipment recognition method. The method comprises the following steps: constructing a power equipment database; pre-training the Faster RCNN network model through a public database; continuously training the pre-trained Faster RCNN network model through a power equipment database to obtain a trained Faster RCNN network model; identifying the equipment picture through the trained Faster RCNN network model to obtain an recognition result. According to the invention, a power equipment database is constructed; pre-training and re-training are carried out on a Faster RCNN network model; the training of the Faster RCNN network model for identifying the power equipment is completed under the condition that the training samples are very limited, the over-fitting phenomenon is effectively prevented, and the recognition precision of the power equipment is improved, so that the abnormal condition in the power scene can be discovered in time, and the potential safety hazard is indirectly reduced.

Description

technical field [0001] The invention belongs to the technical field of electric equipment detection, and in particular relates to an electric equipment identification method. Background technique [0002] In recent years, monitoring electronic equipment has been widely used in various occasions, and a large number of monitoring videos and images need to be processed effectively. In the process of power grid inspection, drones, robots and other acquisition equipment are generally used to monitor multiple power scenes in real time. In order to monitor and process different typical equipment and events, it is necessary to perform object recognition on the collected video images and Accurate positioning. The traditional processing method is to use manual annotation, but the method of manual processing of video images is inefficient and difficult to guarantee accuracy, so there is an urgent need for a target detection technology based on deep neural networks, which can efficient...

AI Technical Summary

Problems solved by technology

However, due to the extremely complex parameters in the convolutional neural network, large-scale training data is often required to train the network. If the data used to train the model is only a small number of labeled pictures, serious overfitting may occur.

Images