Lcd Limited Chinese Character Set Feature Extraction And Recognition Technology

In recent years, digital instruments are widely used in various fields of industrial measurement and control because of its higher precision, more intuitive readings and more comprehensive functional advantages and so on. Although high-end digital instrument already have a wireless communication interface with the computer, in a considerable number of application environments such as scientific experiments, measurement control, meter records, there is still a need for manual meter readings recorded and then transcribed into the computer for treatment which not only is time-consuming and laborious, low efficiency, but accuracy is also poor. Therefore, the digital meter reading has a high application value in the use of image processing and recognition technology automatically .In this paper, digital meter is considered as the research objects to identify a limited set of Chinese characters on a digital type home meter. Watt-hour meter readings include as follows: a positive active general, sharp peak, flat, valley, toward the total reactive power, reverse active general, sharp peak, flat, valley, reverse the total reactive power last month, the need amount of this month, the amount required. The main difficulty in the identification is that the images collected in the system are in natural light conditions, which have lower resolution, much more noise, and lots of size fonts. The research of this paper is as follows:LCD screen binary algorithm: A binary algorithm based on gray-scale morphology is proposed in this paper, the contrast-enhanced image is got by morphology hat transform and low-cap transform, which can filter the uneven illumination due to external interference. Because of no division of segmentation unit, the system avoided the impact from the inappropriate selection of the split cell size. The algorithm proposed is robust and can give a perfect binary result.Correction algorithm: The coordinate axis method is used in this paper. It does not rotate the image directly, but rotates axis angleθand gets the projection of the image in the coordinate axis. Then the distance of the projection between characters is calculated, which can calculate the number of all rotation. Finally, the original image is rotated based on the angle.Recognition of the limited characters : Peripheral features and grid pixel statistical characteristics are combined to identify finite collection characters.