I spent a lot of this week testing the algorithm with with images taken in real time on the Kinect camera. I’m still using my laptop, but should be picking up the Jetson on Monday, I began my first attempt at filtering out duplicate detected edges. I am averaging around 2 – 3 edges with the minimal filtering. I also implemented some image resizing to aid in cup detection from farther distances. This week I plan to get the Kinect working with the Jetson and complete the filtering process.
Team Weekly Status Report (2/28/2021)
A lot of our energy this week was focused early in the week with creating the presentation. We also ordered our parts as soon as we were able to on Monday so that we could get some hardware in our hands. From the presentation feedback we realized that we need more quantitative measurements in our requirements for the launcher and motion planning. We plan to add those specs this week, taking into account ball size, cup diameter, and other metrics.
We were able to all set up the toolchain to build and develop on our github repo on our local machines. We are in the process of getting all of the necessary hardware. With most things set up, we think we are in a good place to get a lot done this week
JP’s Weekly Status Update (2/28/2021)
I created the github repository. Included all dependencies for openCv and the kinect camera. Took a while to get things to build and run properly. To aid in our ellipse detection, I found an open-source ellipse detection package that I can combine with some of my opencv code to make our detection rate more accurate. I was able to strip and reuse some of the algorithm and get it to build in our project. I took pictures of cups on my phone and ran the algorithm to successfully find cup edges when the camera is angled down at the cups. I was able to tune the algorithm to decrease any duplicates around cups but still need to work on filtering out extra ellipse detections sometimes. This coming week I am going to increase the distance that I capture the cup images and retest the algorithm. I am hoping to be able to run the algo in real-time with the kinect cameras instead of saved images.