Application Of Digital Recognition Technique In Handheld Reader

Handheld readers have been extensively applied in lots of systems such as power supply, water supply, gas supply and so on.Nowdays; handheld readers' modes of data input mainly include manual input, RS232/RS485 input, infrared input and wireless input.One of the above modes is sensitive to light or great work intenstiy or has high requests for watt-hours meter.As image recognition techniques have been applying in industrial automation field,at the same time,due to the continuous development of embedded systems,which have a good many properties such as configured,multi-mode communications,convenient network interface and so on,imge recognition meter readers will generated based on image recognition techniques,embedded systems and meter readers,which will get image datas through camera and save restuls of recognition algorithm into readers,which has functions of decreasing work intensity,transmitting or managing datas more convenient and having good graphical interface and will take hold of wide application prospects.The purpose of this paper is that make image recognition technology apply in handheld readers.Under the purpose, some researches have been carried on for the dial plate image location, segmentation and recognition combining with reality and the image processing algorithm and the software interface of dial plate image recongnition have been simulated based on VC++6.0 platform in the paper.The main contents of research are as follows:Firstly,the data acquisition of electic kilo-hours meter,holdheld readers,character recognition technology and developing statuses of embedded systems have been summaried,furthermore,according to the subbject's requests,the embedded Linux development platform has been choosed as the whole image acquisition and recognition application,in the end,the system designs of hardware and software have beed done.Secondly, in view of characteristics of dial plate image, dial plate characters have been automatically recognized through image pretreatment, feature extraction and character recognition.Image pretreatment measurement mainly are as follows:the histogram equalization and neighborhood averaging filtering technology have been utilized to enhance and denoise the dial plate image; the Sobel horizontal operator and Ostu arithmetic have been adopted to locate the dial plate; the hough transition has been used to realize image skew correction.A method has been porposed to locate characters accurately on the basis of sciagraphy and linked list.During the stage of image recognition, according to the structual feature of the dial plater character,one method on the foundation of crossing numbers,horizontal and vertical lines has been improved and holes have been introduced into the character recognition .At last, using BP neural network principles, characters have been recognized after choosing the reasonable parameter and structure.Finaly three mainstream GUI have been analized and the Qt/Embedded interface has been choosed,the cross-compiling environment has been builded for system application and the Qtoipa has been transplanted from PC platform to ARM platform,at last,a simple GUI has been designed based on Qtopia.