Research and Implement Embedded Artificial Intelligence in Low-Power Water Meter Reading Device

This paper presents a system using artificial intelligence deployed on ESP32-Cam to conduct OCR on water meter readings. Data transmission through LoRa technology ensures low-power consumption and long-range data communication. The accuracy of digit classification tasks reaches up to 98%. The lowest current consumption in active and sleep mode is 33.5 mA and 0.2 uA, respectively. With these specifications, the system proposal is proved to be low-power, low-cost, has a long-lasting operating time and can be deployed in widespread use.