Research On The Recognition Algorithm Of Electric Meter

Currently the power sector commonly uses a unique serial number as the identification of an energy meter, and this kind of serial number mostly is replaced by the bar code. While the bar code is added after the energy meter has been manufactured, the power sector has to put up the bar code itself on the surface of the energy meter. However, every meter has a unique table number corresponding to it. So, if can identify the table number of the energy meter correctly though computer vision. Then the table number can be used as the energy meter's identification, so it doesn't need to identify the bar code, even further is that the energy meter doesn't need to add a bar code. This can largely help the power sector to manager their work and reduce their costs. Based on this assumption, this paper studied a recognition algorithm to identify the table number of the energy meter.This paper has briefly described the status of character recognition technology, based on the current research, Regional localization algorithm and number recognition algorithm are in-depth studies, mainly has completed the following tasks:1. Location of the table number of the energy meter: through choosing the proper location flow, mainly including algorithm of edge detection, morphological processing, region restrictions and combining the characteristic of the energy meter itself to achieve the table number region location in the image.2. Recognition of the table number: through extracting the digital coarse mesh features and found that anti-jamming ability of this feature is so poor. So this paper proposed the pixel moments as feature. For the template matching method, it requires a standard template, while the extracted features differ a lot. Use BP neural network as character recognition algorithm. Combining the extracted features from the proposed method, it greatly increased the BP neural network recognition rate.3. Completion of the whole processing: this paper has introduced the image pre-processing, region location, character cutting, feature extraction, character recognition in detail, has realized the whole processing, a lot of experiments were conducted to validate the method adopted in this paper.Through doing experiments to the collected images, experiment results show that this method can better achieve the region location of the energy meter and table number recognition, with a certain robustness, the accuracy rate of the table number can reached 87.8% and the table number of the image recognition rate can reach 96.8%.