CNN-LSTM-based temperature instrument number identification method and device

The invention provides a CNN-LSTM-based temperature instrument digital identification method and device. According to the method, modeling is carried out by combining a convolutional neural network with a long-term and short-term memory network in a recurrent neural network, so that complicated preprocessing processes such as excessive cutting do not need to be carried out on picture data in an input model, prediction can be carried out by directly taking a reading picture as an integral input model, and the preprocessing process of reading identification is greatly simplified. The long-termand short-term memory network part adopts a bidirectional long-term and short-term memory network, and meanwhile, past and future information is considered, so that a prediction result is relatively better in performance. A connection time sequence classifier is adopted to decode an output sequence, the problem that input and output are difficult to correspond is solved, many steps are simplified,and the prediction efficie