Mutual inductor nameplate text area detection method based on deep learning

The invention discloses a mutual inductor nameplate text area detection method based on deep learning. According to the method, a first-stage model is used for detecting a text area on a nameplate of the transformer equipment by using an image pixel classification principle. The mutual inductor nameplate image feature extraction and fusion method adopts a U-Net network multi-dimensional feature fusion method, and features of character areas with different sizes in an image can be accurately extracted through the method. Meanwhile, in order to improve the recognition performance of the long text in the transformer nameplate image, a Difference Binarization (DB) network is adopted to associate, map and classify the fused features in the text detection stage, so that the situation that the long text with semantic association is cut off during text detection is avoided. Therefore, through a mode of combining the U-Net network and the DB network, the detection capability of the model on small-region texts is impro