Week 4

 


 

This week showed that most of our efforts on this project paid off, resulting in successful tracking of objects that are displayed in front of the camera.

Within our system we now accomplished the following this week:

  • -          The Raspberry Pi now has the appropriate power supply and so does not shut down any more, allowing the tracking to work without low voltage problems.
  • -          Automatic movement of the Pan-Tilt mechanical arm to follow object that is being tracked. In the video above we can see the system is tracking the phone being used to record it.
  • -          Live image feed with labels showing each object and what it is

-           

We still however ran into the following problem:

  • -          The Raspberry Pi 400 arrived but unfortunately after removing the internal Raspberry Pi 4, it was discovered that it does not contain a place for the camera module to be connected to. Therefore we continued working with the Raspberry Pi 3B and looked for alternative Raspberry Pi system that we could use to increase our performance.
  • -          Due to hardware limitations the object tracking is running at roughly less than 2 FPS and so when tracking the object, it sends commands to the Pan-Tilt after each image processing is done, meaning that the Pan-Tilt moves less often and in big jumps so the tracking is done slowly and if the object is moved quickly off the screen, it will no longer be tracked by the system. It can only track slow moving objects with this current hardware, but once we obtain a newer Raspberry Pi and the coral USB accelerator, we expect the system to run at roughly 25-30fps, meaning that the images will be processed much faster and the pan-tilt will be able to move fluidly, keeping the object in view even when moved quickly.
  • -          When an object is displayed in poor lighting conditions, it is not accurately detected by the system. This can be fixed by soldering and connecting the NeoPixel LED to the Pan-Tilt to allow a constant light source to allow accurate detection of objects.
  • -          Currently the system detects all objects that are displayed in front of it. We would like to add a feature that will allow us to set it so that only one object will be tracked and displayed on the screen. So that even if it goes off the screen completely and then is brought back later, the system knows to track it.

Tasks for next week:

  • -          Solder the NeoPixel LED to the Pan-Tilt mechanical arm so that it provides a constant light source required for accurate object detection.
  • -          Continue looking for a better alternative to the Raspberry Pi 3B we currently have.
  • -          Add the ability to only display, identify and track one object to the system.

Comments

Post a Comment

Popular posts from this blog

Week 2

Image
  This week, we accomplished the following: -           Found a viable power source and display cable. We used a laptop to power the Raspberry Pi. -           Connected a Raspberry Pi to the computer (LEDs lit up to show connection). -      Connected the camera to the Raspberry Pi and tested to see it was working correctly. -      Assembled the Pimoroni Pan-Tilt HAT and mounted the camera in to the Pan-Tilt -           Found resources to help with the software side of the project, enabling and taking pictures with the camera.   However, we ran into the following problems: -           Raspberry Pi and Coral USB accelerator ordered did not arrive in time, they were backlocked. So we had to use a much older Raspberry Pi 3b that was brought as a contingency...
Read more

Week 3

Image
  This week was one of the more productive weeks as we managed to: -           Connect and run the Operating System of a working Raspberry Pi. -           Connect the camera and the Pan-Tilt mount to the Pi board. -           Use researched code to operate the camera and take pictures with it. -           Connected the wiring and get the Pan-Tilt to be recognised by the Raspberry Pi. -           Produce code to detect and move the Pan-Tilt mechanical arm using the arrow keys. -           Get object detection working with the system. The system is able to detect objects that are shown in front of the camera and display what it thinks the object is with the percentage showing how accurate it is in real time. As sh...
Read more