Based on visual model and LSTM cyclic convolutional neural network Automatic recognition method for instrument readings

In order to achieve automatic acquisition of instrument readings, a recognition method based on visual model and Long Short Term Memory Network (LSTM) is proposed. This method introduces an attention mechanism model to find the location of the feature that has the greatest impact on the final prediction result in the feature map. Once the dial position is determined, this method utilizes Long Short Term Memory Network (LSTM) to recognize text within that position. The experimental results show that compared to traditional methods, this method avoids the character cutting and a large number of manual modeling operations required by traditional pointer reading recognition methods, significantly improving recognition accuracy.