A large part of this past week was spent looking through old blog entries, photos, and data to come up with material for my upcoming presentation to the Raspberry Pint club (see
https://www.eventbrite.com/e/raspberry-pint-raspberry-pi-and-other-digital-making-fun-tickets-80736249295 for tickets and details). I read on the club's Facebook page that something like 100 people have signed up for the event so far, so it should be some great exposure for AMOS!
It was interesting to look back on all the work that was done for visual image processing for the purpose of object detection. None of this work (which consumed a couple months of my time) proved to be very reliable for finding objects, but some of the generated camera images were kind of cool:
The above photo was taken by the original AMOS in July 2018. A 640x480 color image from the Pi camera was processed using the Canny Edge Detection method. At the bottom of the image you can see the edge of the boat's hull, and in the middle you can make out a person paddling a small kayak in front of a bridge. The trouble with all of the visual object detection stuff though was that the appearance of the water was too variable. Too many waves, reflections, and nuances in color that made it very difficult to use an algorithm to find non-water objects.
As a bit of a distraction from the presentation stuff, I also continued plugging away at the data visualization software for AMOS. There is nothing ready to show there yet, which reminds me of one of my all-time favorite Dilbert cartoons:
No comments:
Post a Comment