Gas-meter Reading System Based On Digit Image Recognition And Compressed Transmission

As social information is increasingly developed, gas meter's automatic reading system has become an important part of public service management industry. However, existing gas meter records the usage of gas by means of pulse accumulation, it needs to transmit every pulse signal and doesn't make mistakes during communication transmission. At the same time, present automatic reading system can't check gas meter remotely. Contrary, it needs people to check gas meter personally. Aimed on above question, the gas meter reading system based on digital recognition and compression transmission is presented and designed in the paper. And this system achieves a series of function of meter reading based on digit image, transmission and digit check by ways of retrofitting cheap cameras and transmission devices in the existing meter.First of all, overall system frame is designed according to function needed to be completed in this paper. According to needs, the systems include image acquisition module, image's digital recognition, and compression transmission. Image acquisition module is aimed to collect original image, make use of recognition algorithm to compute binary, then compress and transmit middle image, finally store date base and recognize digit in the backstage management systems.The key algorithm of the system is digit recognition based on image. A series of digit recognition algorithm of image pre-processing feature extraction and sorting algorithm is researched in this paper. First of all, in image pre-processing aspects, aimed at the characteristics of table character of gas meter, it exploits a series of algorithm, such as gray-scale enhancement and skew correction, to do pre-process initial data, then combined with projection methods to segment digital character. Finally get a identify numeric characters of five bit rules. From the perspectives of feature extraction, a robustness alphabetic string featured extraction algorithm is presented. Based on original feature, modified line number feature extraction algorithm is come up and increase diagonal line parameter of four digit Image, which can show the difference of similar number. In addition, the recognition of Incomplete half word character is further researched. In the sorting algorithm, modified PSO-SVM parameter, Optimization Classification Method is used and improve its precision of recognition. In the compression transmission, run length coding and Huffman coding is exploited so as to compress and transmit binary Image Information.Finally, platform build test is implemented in the paper. Digit recognition test is carried out without changing original structure. Results show that recognition accuracy rate of complete digit has reached 96.10%, and recognition accuracy rate of incomplete digit has reached 88.90%. Compressed Image Transmission is tested. On the basis of without changing recognition accuracy, the rate of compression reaches 18.33:1,which extremely reduce the burden of communication. Result shows that the system designed in this paper has a good accuracy and stability.