.png)
Python, Kivy, OpenCV, Android, Buildozer, RaspberryPi, Arduino, Chromebook etc.
In the beginning of the pandemic, I guessed that the forehead temperature measurement of each employee or customer would be a standard until everything goes back to normal. So I thought, maybe I can act fast and start up a small, local business. But so many things went wrong even to develop a prototype, because of a little bit of bad luck, a little bit of inexperience. I tried many things and I would like to share story of incomplete project with you.
Scope: Measure employee's temperature to catch if one has fever, a COVID-19 Symptom.
Idea:
To prevent handheld measurement to save time and reduce contact.
Ensure each employee is measured three times a day.
Employees can measure themselves
Either recognize employees with face recognition or RFID tag card.
Employees stand in front of the device and get measured
Can be like a station or something to hang on the wall
Can be sold as a system or with monthly subscription
And here are triple constraints of my project as PMI defines:
Scope: Easy, I want to stick on my scope. Though, cost and time will affect it.
Time: I have to be quick, because I am sure this is not only my idea. Even though, it was not done before (AFAIS).
Cost: Oh yeah! Am I planning to invest on it, or be cheap :/ (Something I learned here)
>>> Quality: With all these, can I take out a product under quality standards of a commercial product?
Plan 1: Android app
At first I decided to design this on an android app so I could use various tablets. I could design a compact sensor module in a plastic casing that is pluggable either to tablets usb port or headphone jack. However, I didn't know how to develop an android app which can process image and interact with a database. I checked with my peer environment if I can find someone to be a partner. Then, I couldn't find anybody. So I decided to change plans.
PLAN 2: Raspberry Pi and Smart Mirror
If I was going to be alone in this project, I should have gone with what I know best. So I considered using a Raspberry Pi so I could write my software in Python and face-recognition python module that I experienced before. Please, check out this useful library. However, I still needed to cut the cost and an LCD is not very cheap.
My plan was to do something useful, less costly and good looking. Instead of using a big LCD, I could do something like a smart mirror and reduce the lcd size. So people can also see their face and fit their face in an area for measurement. This is what was on my mind.
It was costing me around $140
Materials:
I have never built the whole system because my problem was much bigger.
There was literally no sensor in stocks of any store. Digikey, Mouser, Adafruit etc. All zero in stocks. I could only find these and I didn't even know if any of these would work.
TPIS 1T 1084 has I2C and ZTP-115M has analog outputs. I connected and tried them on Arduino. I couldn't make any of them work. I failed on TPIS probably because of I2C Voltage or somehow I damaged it. Anyway, it was not something useful for this project since it has +/-3 C tolerance. And at least from ZTP-115M I could read varying voltage output, however I could not manage to do correct mapping. Actually this sensor is mostly used for industrial purposes.
Then, I kind of gave up. But at least I wanted to do face recognition on Android side. Because, It would be useful for me to have it for some possible future projects. At the same time, I stopped going to work and started to work remotely. Quarantine started..
In part II, I will explain how I struggled with Kivy on Android and how I tried to reverse engineer a sensor of an off the shelf handheld temp device. It was getting more interesting.