Group Actions
Engagement
  • Author Author: dubbie
  • Date Created: Date Created
  • Views 3037 views
  • Likes 9 likes
  • Comments 27 comments
Related
Recommended

CatDogFoxBot #5 : Collecting Some Temperature Data

dubbie

The GridEye sensor is working so now I have to work out how to recognise when a cat comes along (or a dog or a fox). To help with my own understanding of what the GridEye sensor data will look like I have set up a simple experiment of the GridEye and a boiled kettle of water in an otherwise unobstructed area, but inside the house. I have changed the data displayed so that only temperature values higher than the average are displayed with anything else just being blanks. This is to help 'see' only those elements that are hotter than the background. I realise that this doesn't really work for large hot objects which fill the sensor point of view but as the sensor angle is 60 degrees this isn't going to be a problem except at very short distances.

 

So below is the data obtained at 50 cm

 

image

 

This is the data at 75 cm

 

image

 

This is the data at 100 cm

 

image

 

This is the data at 150 cm.

 

image

 

This is not quite what I was hoping for. At the close range there is a nice cluster of the higher temperature pixel values but  as the distance increases the number of pixels registering the higher temperature reduces quite quickly, which I did expect, but also the temperature drops as well, which I was not expecting. But after thinking about it, it does make sense. The sensor has a wide 'beam width' of 60 degrees so the area collected by each sensor increase rapidly and the temperature shown is going to be an average of the whole of that pixel area. So the sensed temperature will drop as more and more of the non-object, or background temperature, is included. This can be seen most clearly in the data at 150 cm where there is almost no discernible higher temperature pixels, just one pixel at 24 degrees, with the rest effectively being the background temperature.

 

This data is collected inside the house on a relatively warm day using a pot of hot water and now I will attempt to collect some data from outside during the early evening into the night to see if it might be better. At present I do not have a long lasting battery power source so I will be running the system from a USB cable out of my office window. I know that cats sometimes walk past this window so perhaps they will oblige tonight.

 

There is also the problem that cat fur is designed to keep cats warm so although the cat body will be at a higher temperature of approximately 38 degrees, the external temperature of a cat might not be. Maybe I'll find out tonight.

 

Dubbie

Parents

  • Dubbie,

     

    The results are quite disappointing, but still understandable given the wide beam width.  Sure, you might be able to make it work, but might take quite a bit of effort to do so.  Your comment of cat fur got me thinking so I pulled out my IR thermometer and shot a couple of readings.  Of course my cat was near by and as long as I let him chase the laser dot, he was game to help me collect some data.  My hand ~34C, his body (black or white fur) ~28-30C, while inside his ear is ~36C.  I hope this helps.

     

    Good luck with finding a good solution for your project.

     

    Gene

  • in reply to genebren

    Gene,

     

    Thanks for the cat temperature information. I was wondering how I might get hold of a cat to find out. If the cat is at 30 C then it should be possible to detect it at night against a colder background. Possibly not during the day.

     

    The 60 degree beam width is more of a problem than I was thinking when I started. Still, the system does work. I ran it for several hours last night but did not seem to detect anything cat like. I was surprised by how much the background temperature does seem to change from second to second, by two or 3 degrees. I expect it is something due to wind.

     

    I need to distinguish localised hot spots from slowly changing general background temperature changes. I don't think using absolute temperatures is going to work so my current thought is to normalise the data and then maybe use an artificial neural network to process the data. There is so much data I need something to flag up interesting data. I might also put the sensor across a path where I know cats walk as well.

     

    Dubbie

  • in reply to dubbie

    I think perhaps has the answer in his video here:

    Building a Thermal Imaging System with the Raspberry Pi 3, Sense HAT and Panasonic Grid-EYE

    ... just need to get the cats to carry candles, and you are sorted... image

     

    There was also a road test of the Grid-Eye sensor here which contained some other tests which may be of interest:

    Panasonic Grid-EYE RoadTest - Review

     

    I also seem to recall that was doing something with monitoring cats outdoors on Hackaday, so may be worth a look for some ideas there:

    https://hackaday.io/project/10472-cats-in-motion

     

    I have found that the PIR sensors in household security lights will pick up cats (and even hedgehogs), so perhaps can add one or two on as additional sensor data.

  • in reply to dubbie

    Dubbie,

     

    In thinking about your issue with the 60 degree beam width, I was wondering if you might be able to use a pin hole aperture in front of the sensor (similar to a simple pin hole camera) to reduce the beam width.  There should be plenty of articles out there that would help with focus distance and such.  Just a thought.

     

    On the topic of how to handle the data from your grid eye sensor, there are a lot of image processing topics that might help you break the data down into small chucks.  I like the ideal of normalizing data.  This is a good starting point.  From there you can segment objects (adjacent pixels) into connected blobs.  These objects could then be tracked to determine if they move or change.

     

    Best of luck!

     

    Gene

Comment
  • in reply to dubbie

    Dubbie,

     

    In thinking about your issue with the 60 degree beam width, I was wondering if you might be able to use a pin hole aperture in front of the sensor (similar to a simple pin hole camera) to reduce the beam width.  There should be plenty of articles out there that would help with focus distance and such.  Just a thought.

     

    On the topic of how to handle the data from your grid eye sensor, there are a lot of image processing topics that might help you break the data down into small chucks.  I like the ideal of normalizing data.  This is a good starting point.  From there you can segment objects (adjacent pixels) into connected blobs.  These objects could then be tracked to determine if they move or change.

     

    Best of luck!

     

    Gene

Children