Hardware Overview

This section discusses some of the design choices we made in terms of the hardware

Brains

The Nvidia Jetson-Nano was chosen as the brains for this project because of the amount of power this device has. An arduino was obviously not an option and we were unsure if a raspberry pi would provide enough processing power to achieve our goal of real time vision processing. So we went with the next step up after the raspberry pi, the Nvidia Jetson-Nano. We purchased ours for $75 off of ebay. The Jetson-Nano comes equipped with a quad core cpu @ 1.5GHZ, 128 cuda cores, and 4GB of memory - the cuda cores were the most exciting aspect about this device for us because now we could utilize both the cpu and gpu to do our processing (See the Cuda Acceleration section for more details).

Eyes

The eyes of this project is one CSI raspberry pi camera v2. This camera supports a maximum fps of 120fps @ 720p but could be pushed further if you did some software tweaking, we were satisfied with 120fps. The maximum resolution is can record is up to 4K at 15fps. Our camera was purchased for $25 from MicroCenter. We also needed a CSI cable to connect the camera to the Jetson-nano, so we purchased a 1m CSI cable from MicroCenter as well for $5.

Arm

using the FOV of the camera, the exact height the camera needed to be above the table was calculated to be 52in. An arm was constructed out of 1/2 box section to hold the camera parallel to the game field at the exact location.