Abstract：The emergency of Narrow Band Internet of Things （NB-IoT） provides an effective solution to the problems of difficulties of meter reading and management in existing meter-reading systems.However，the existing NB-IoT-based meter reading systems usually have high power consumption，high installation cost and limited application scenarios，which obstruct their popularization.In this paper，we propose MeterEye，a general remote meter-reading system based on NB-IoT and image processing.MeterEye avoids the modification of the original weter by reading the dial picture，and only transmits the amount of image change containing the critical information which is greatly compressed by compressive sensing to reduce the transmission power consumption.In this paper，the quality of image reconstruction is analyzed for both the pointer and wheel meters，and the accuracy of the digital recognition system based on Convolutional Neural Network （CNN） is used as the standard for measuring the quality of image reconstruction.Experiments show that when the compression rate is larger than 7%，the digital recognition accuracy of image by CNN is higher than 94.09%，and the transmission power consumption can be reduced to 15.99%.
王杜毅,常相茂. MeterEye：基于NB-IoT和图像处理的普适远程抄表系统[J]. 小型微型计算机系统, 2021, 42(9): 1950-1954.
WANG Du-yi,CHANG Xiang-mao. MeterEye：a General Remote Meter Reading System Based on NB-IoT and Image Processing. Journal of Chinese Computer Systems, 2021, 42(9): 1950-1954.
2021, Vol. 42 
