I tried the GridEye sensor outside last night where the temperature dropped to about 10 C, which is actually better for detecting the cats as they would stand out more from the background ambient temperature. I did have to make a few tweaks to the settings and the training data set for the Artificial Neural Network as all my previous testing had been done inside where the temperature is about 20 C.
Regretfully, no cats were detected last night after several hours of sensing. I think it might be because the passage along side my house where I put the detector is pretty dark and although cats might be good I don't think they will jump down from my back gate into a darkened passage way. I had only seen cats there during the day, so on thinking about it, it is not surprising I didn't detect anything. The cats tend to stick to the far end of my garden and I was not able to set up the GridEye sensor system there (I'm not going to leave my laptop out all night!)
Time has now run out but I thought I would try the system inside and try to detect my legs as I walk by. I've used the same training data for the ANN but tweaked the normalisation values to compensate for being used inside. I also increased the sampling rate to about once per second as it is quite easy to walk past the sensor faster than the previous slower sampling rate - which might be another reason why no cats were detected. The graph below shows the data obtained from walking past the sensor, turning around, waiting a few seconds and then walking back.
There are two clear peaks, the left most one is walking past the sensor and the second is afterturning around and walking back. It looks as if it might be eminently possible to detect cats using this sensor (if only I had a trained cat to use for tweaking the normalisation values and ANN training data sets). It is quite interesting to note that there seems to be some transitory effect from walking in front of the sensor where the background level doesn't immediately return to ambient. I noticed this effect last night when setting up the sensor as I used myself as the test subject and walked backwards and forwards in front of the sensor. There seems to be a slight continuing temperature increase after each walk past which takes a few seconds to return to ambient. It is quite likely that this is a swirl of warmer air slightly heated by my body that takes some time to dissipate. I saw this effect at the Tate Modern Art gallery in London UK once, where there was a large rubber balloon floating about in the air. It was so delicately balanced that just walking under it caused it to rapidly rise, simply due to the warmer air rising from a human body. It was quite a fun thing.
Time for this Project14 competition has now run out so no more time to try looking for cats tonight. Plus it is raining a lot and I'm not going to put the sensor outside. So this is the end of my CatDogFox robot detector system. Although it looks like it should work the lack of a trained cat means I might never know. I didn't manage to get the mobile robot part I had planned, implemented, nor the weatherproof casing, or scanning system. Still, it was fun playing with the GridEye sensor, as well as the Artificial Neural Network, as well as the MKR 1000.