On Automating a Lux Meter Reading

Published on 2020-03-25

A requirement came up for a project I've been working on recently to continuosly collect lux values+images on certain surfaces. Since the data collection needed to be done for a considerable time period manual collection was not feasible. To automate the collection of lux values, bellow ideas were considered.

  1. Use the mobile phone lux meters, do a mapping from those numbers to actual lux meter numbers. Requires writing/ finding app to collect the lux values periodically. Don't know about the accuracy of lux meter in mobile phone. Need to align that values to pictures from ip cam.

  2. Place the lux meters so the ip cam see the display clearly, and read them from image using ocr. This is the ideal since no need to maintain an alignment form ip cam image to seperate lux reading. Unfortunately the ip cam resolution is not enough to read the lux meter display clearly. Even ip cam is zoomed in it's not reliable(and if you zoom in you loose scenario data).

  3. Setup a mobile phone near to lux meter so it can get a picture. and do ocr on it. Still need to align 2 images( ip cam and moble phone cam). Need a mechanism to set up the mobile phone next to lux meter. Anyway among the optons we have this is the most feasible so far.

  4. Buy lux meters with transmitting feature.

After doing a feasibilty study decided to go with option 3.

First placed the mobile device slightly above the lux meter display. Then installed an ip-cam app in the mobile device to make the data collection easier. To sync up the images(in time dimenssion) used the time stamps stamped by ip cam and mobile app. Synced up the time in ip-cam and mobile device to same time. Then using an ocr script read the ocr value from image and anotated the ip-cam image. To read the images first tried tesseract. But found out it's more suitable to read the printed stuff(our lux meter has seven segment display digits). Then we found a way which utilized a mechanism based on contrasting area coordinates. after few tweakings it gave us right readings.

Results

input image lux reading

Unfortunately this setup in neither portable nor scalable.