1. Project # 12 – Smart Walker REU student: Christopher Feltner Graduate mentors: Sharare Zethabian and Siavash Khodadadeh Faculty mentor(s): Dr. Turgut and Dr. Bölöni
Week 2 (May 21 – May 25, 2018)
Accomplishments:
Literature review, literature review, and more literature review!
Rewrote and reorganized Smart Walker code using object-oriented design principles. These updates to the software will make it more extensible and robust. Plus I squashed some bugs along the way.
Put code through peer-review and implemented suggested changes
Problem & Solutions
Jon and I are concerned that the Raspberry Pi will not be powerful enough to run CV in real time from the Kinect
Issues: USB 2.0 connection on the Pi is too slow to handle both the depth and RGB images produced by the Kinect; not enough processing power for real-time object recognition using neural network
Possible solutions: throttling frames-per-second; switching between input streams so that they are used only when necessary; changing hardware; changing obstacle recognition algorithm
Requires further investigation
Plans for next week:
Figure out a path forward for CV – further investigation
Begin to write software to recognize obstacles in images and collect distance information
NSF REU Research Experience on Internet of Things 2018