Design and Research of Unconnected Point Numeric Recognition System Based on Neural Network

Currently few people focus on the research of printed unconnected point numeric recognition, this paper, extract this kind of number from the table number of electric energy meter (EEM). We have designed a system to recognize this kind of numeric based on neural network. Using the normalized unconnected point number image (28x20) as the neural network input. Experiment results show that the system can better achieve unconnected point digit recognition, with a higher recognition rate, even deal with relatively poor picture; recognition rate can be maintained average above 96%.