Research on Fast Numerical Recognition for Numeral METE

Traditional digital display instrument lacks the communication interface, and its intelligence is based on digital identification technology. To meet the real-time requirements of digital identification, this paper puts forward a complete set of digital recognition algorithm. It uses the Otsu method to carry out the threshold segmentation algorithm on images to be identified. Then it uses projection method to determine the character display area and single character segmentation. At the end, character recognition is accomplished by the method of strokes coding. The process doesn´t include traditional steps like refinement, morphological computation and neural network training, which bringing high computing speed. Experiments show that this method has better adaptability to rotation, scaling, illumination changes of targets to be recognized, meeting the time and accuracy requirements in digital instruments.