Thermal Cam for COVID-19

We would like to introduce our computer vision body temperature measurement system.

There are a lot of body temp thermometer however still one by one and although no contact infrared measurement but a few cm form the head.

When medical staff contact many patients to measure their body temperature, there are still many risk for them.

We have made our best solution for less risk and the lowest cost as possible.

Our solution make less risk for Medical staff to measure body temperature for many people but low cost.

This sensor is also affected by distance ±0.1℃/10cm.

So we ask one stand just in front of wall to keep the distance same to measure.

We are using Microborometer sensor(FLIR Lepton).

It must be Lepton 3.5 with Radiometry

Thermal image is transmitting via Raspberry Pi 3+ via WiFi&LAN to iPhone.

Thermal Cam New Version !

We are there at FLIR Lepton Developer page.

This system is also required iPhone app Thermal cam 3X with iPhone11, 11Pro or iPad Pro(2020) to run our image recognition by A13 Bionic or A12X.

Image recognition run by A13 Bionic and thermal image transmits from Raspberry Pi3+ to iPhone via your local router.

We are developing thermography system with FLIR Lepton sensor.

To start with We developed small thermography(Microborometer) sensor on toy drone with iPhone tethering as per following video.

And with this system, we maintenance Solor panels in our sight.

Later we combined object detection to find error panels with NVDIA Jetson.

• This time not Jeston solution, sorry but not much time left for us. Jetson Nano is too slow and too late to developed for us for Jetson NX released.

Now you can understand how we have developed our thermal system with deep learning.

we have developed this COVID-19 detection system as following video.

There have been 2 issues for this project for us.

One is sensoraccuracy and the other is cost.

Microborometer is not suitable to measure the absolute temperature.

Tolerance is roughly ±3℃ differential but Relative temperature is 0.05mk(℃.

The absolute temperature is floating as above.

We fix the absolute temperature using black body furnace.

And offset lepton raw data based on M+(memory temp) in our app.

It means thermal image is floating like this point could.

With black body furnace. it will be the best solution and accuracy.

However black body furnace costs more than $1000/unit.

If you measure body temperature every 10-20 minutes in the well-airconditioned room, it will not be far from real body temperature but every time you stand where the suspect person, which none want to.

We found wall temperature is still stable for some time.

One of our better solution for the accuracy this time. with this solution, you can measure body temperature for 20-30min feasibly with our system.

For our apps and installation for image recognition, pleas study our following web page.

https://kansaidenki.biz/new-thermal-cam-3x-b/

https://takesan.hatenablog.jp

And refer all of our information, which will be required to take your whole day to install Raspbrrey Pi3 with some skill for Linux.

For your better understanding our Rough overall idea is as follows;-

We study our room wall with well air-conditioned and we measure wall temp with our temp logger(refer to following chart) and found it one of solution without black body furnace.

Well air-conditioned room temperature is stable as above. especially every calibration for 10 to 20 minutes, you do not have to use black body furnace.

our internal checking as attached.