Experimental validation demonstrates significant reductions in OCR processing time and server load

compared to conventional full-frame pipelines, while maintaining reliable digit recognition accuracy. The

complete hardware design, firmware, and documentation are released under an open-source license to enable

reproducibility and large-scale deployment.

Analog water meters remain widely deployed due to their robustness and regulatory approval; however, their

lack of digital interfaces necessitates manual readings, resulting in delayed consumption analysis and operational

inefficiencies. This article presents an open-source, low-cost hardware platform for retrofitting existing analog

water meters using an ESP32-CAM imaging node combined with a custom power and illumination PCB. The

system captures periodic images of the meter dial and transmits them to a centralized server for optical character

recognition (OCR).

Experimental validation demonstrates significant reductions in OCR processing time and server load com-

pared to conventional full-frame pipelines, while maintaining reliable digit recognition accuracy. The complete

hardware design, firmware, and documentation are released under an open-source license to enable reproducibility

and large-scale deployment.