Measuring the utility consumption of end users is one of the principal processes for every utility provider. Yet, the technology for “smart” electricity, gas, water or heat consumption metering is still immature and their rollout is therefore slow. The meters must be changed one by one - a process that would take years even in the most tech savvy utility companies - until the transition from the traditional analog meters would be complete.
Until then, companies have to read those analog meters: a costly task with lots of unnecessary labor. Noticing the digits, writing down the consumption and meter ID by hand, transferring it into .xlsx sheets, feeding it to CRM systems; a slow, cumbersome process prone to error and subject to fraud.
Together with the partner company ISDD, Cognexa analyzed the workflow of various meter-reading crews to develop a mobile application that makes the entire process chain exponentially more efficient and convenient.
The previously cumbersome chain of tasks got reduced to one single and easy step, namely taking a picture. Our OCR technology locates the meter display, reads the digits and sends them directly to the utility provider’s CRM system once Internet connection is available.
The reading itself is carried out offline, which solves the problem of poor Internet signal in the underground pits. Fraud prevention mechanisms include analysis of GPS location and a timestamp of the photo, plus a reference point containing each meter’s past values. The application automatically alerts the measurer, if the photo is too blurred or otherwise unreadable.
The application is still under development and is to be deployed in the first quarter of 2018. Our client expects an increase in the work efficiency of the crews and overall higher satisfaction, due to a smoother process. Given the archiving and automation of the measurement verification, trust between all parties involved will increase as well.
Offline mobile deep learning app for seamless water meter reading
3 months to fully functional Android app
TensorFlow, Python, Java, Android, CxFlow, OpenCV