Automatic reading of domestic electric meter: an intelligent device based on image processing and ZigBee/Ethernet communication

摘要

In undeveloped areas around the world, many traditional meters need to be upgraded. Compared with replacing the mounted meters with high-cost modern ones, it is a better choice to upgrade them with new technologies. In this paper, an automatic reading system of the traditional household meter is designed on the basis of image processing and advanced DSP system. To identify the meter reading accurately, a regional average method is proposed to implement the image-scaling in order to avoid the distortion. In the image-filtering process, we raise an average-product method which is verified to attain good effects. For image segmentation, a new union thresholding method, based on the grayscale transformation, is proposed to enhance the adaptability of uneven luminance. Iterative rejection is applied to decrease the errors during character localization. Then, a training sample library with 1,400 characters is designed and collected for the training of the BP neural network. For data transmission, NAT technology is introduced to build data connection between the remote server and the data collectors working in the local area network. According to the field test, the proposed system can obtain a recognition rate of 99.7 % under normal environment, with the identification period below 2 s, while the resulting data can be transferred reliably through 1-3 walls in ordinary buildings.