Summary. As mentioned in the team status report, we had some setbacks this week regarding the integration of the system caused by some hardware issues with the communications chips. I had planned to have begun sending object data from the Jetson to the Arduino Due MIDI controller by the end of this week, but due to these issues, I will need to do this next week. Instead, I chose to address an issue with the contact detection subsystem I had identified last week, which was that the glove markers were blurry and impossible to detect when the image wasn’t still. This was a critical flaw, because it meant that the user needed to restrict the movement of the camera and their hand to accurately detect objects, which would make the device hard to use. I spent most of the week researching the issue and fixing it.
Removal of motion blur. I considered two ways of resolving the issue. The first was to add more logic to the glove-tracking procedure that could be resistant to the effects of motion blur; for example, I considered using an optical flow technique called Lucas-Kanade Tracking to maintain a best guess of the glove’s position between subsequent detection’s of the fiducial marker. The other approach was sensor-based and involved adjusting the technical parameters of the camera to eliminate the motion blur altogether.
I chose the latter approach for simplicity and spent the week learning how to raise the shutter speed (lower exposure) while also increasing the brightness of the camera. I ended up doing this through the v4l2 Linux device driver. I experimented with different values for these parameters that would strike a good balance between image sharpness (which increases with lower exposure/higher shutter speed) and image brightness (which decreases with lower exposure).
Schedule. Due to the setback with board-to-board communications, we are slightly behind schedule, as mentioned above. This week our schedule calls for implementing and verifying overall integration of the system. This should still be doable, but will require more work and closer communication between our team to meet this target. I will work with Harry to get data off of the Jetson and onto the Due. I will also work with Kevin on the glove design and assembly; this will involve creating an enclosure for the Arduino Micro, and creating and attaching more durable fiducial markers onto the front and back of the glove. Personally, I will begin working on improving the display graphics to be more intuitive.