I somehow mounted the camera array and the Raspberry Pi above the cooktop (you don't want to see any pictures, believe me) and started the measurement. Today my family planed to have weißwurst (https://en.wikipedia.org/wiki/Weisswurst) for brunch and I used the change. These sausages are already cooked, so you only have to heat them up to have a warm meal. On the other hand you have to avoid to boil them because then the sausages will explode. So heating them up is a little bit complicated if you want your sausages hot. Perfect for the stove assistant!
Here is the video. It is 15 times faster than in reality.
The stove assistant also saved the temperature profile (all values in degrees celsius):
For the first test I didn't make any complex image recognition and temperature detection. I simply took the highest measured of the grid-eye sensor and assumed that this is the temperature in my pot. As long as there is only one pot at the cooktop this should be OK.
There are some interesting findings:
The spikes in the temperature profile always accured when I stirred with a spoon. So I assume the temperature at the surface (which the grid-eye sensor can measure) is about 10 degrees celsius colder than the actual temperature in the pot.
It seemed that the water was nearly boiling although the sensor measured only 60 oder (when stirred) 70 degrees celsius.
So I definitely have to make more tests and measurements.
Nevertheless we managed to heat up the sausages corectly:
more measurements ...
And a new camera mount above the cooktop and start with the HMI.