Abstract:Aiming at LED display meter, put forward a character recognition method. Its character segmentation applied edge detection, line detection and gray projection to rapidly extract target character. It recognized number and decimal point separately and improves the self-adaptive threshold number recognition based on the statistical feature of numerical strikes. Experimental result indicates that the accuracy of the proposed recognition algorithm reaches over 95% and the time-consuming is short, which demonstrates the algorithm of high accuracy and good real-time.