Image recognition method for secondary circuit terminal based on contrastive learning and improved CRNN

An image recognition method for a secondary circuit terminal based on contrastive learning and an improved CRNN, comprising:

• step 1:
• pre-training, by a server, sample data of a secondary circuit terminal block of a power system captured by using on-site equipment through the contrastive learning;
• step 2:
• improving, by a server, a feature extraction layer of a Convolutional Recurrent Neural Network (CRNN) by using a residual neural network;
• and
• step 3:
• introducing, by a server, an Efficient Channel Attention Network (ECA-Net) on the basis of the step 2 to construct a recognition model for the secondary circuit terminal based on the contrastive learning and the improved CRNN, wherein the recognition model for the secondary circuit terminal based on the contrastive learning and the improved CRNN is configured to recognize a secondary circuit terminal to be recognized captured by using on-site equipment,
• wherein the step 1 comprises:
• data augmentation:
• expanding data, by applying cropping, flipping, rotation, random noise adding, random Gaussian blur and color transformation techniques to unlabeled data, wherein data expanded from same-source data are used as similar sample pairs with each other, and remaining data are used as different sample pairs;
• encoder network:
• mapping the similar sample pairs and the different sample pairs to a feature space, extracting a feature capable of distinguishing samples and similarity, and outputting a high-dimensional feature vector through an average pooling layer;
• and
• projection network:
• projecting the high-dimensional feature vector output by the encoder network to a low-dimensional space to obtain a feature vector output by the projection network.