Step 1: The Single Board Computer (SBC) - Hamidreza-Alai/Vision-Based-Vehicle-Tracking-System GitHub Wiki
A) Introduction
This page deals with "deciding about the single board computer" for the vision-based vehicle tracking system project. Important factors are:
-
OS: The factor of possible Operating Systems (OS) to be installed on the computer is very important. Preferably, it is desired to install a Linux operating system such as Ubuntu or Debian.
-
CPU: A powerful CPU is preferred.
-
RAM
-
USB Ports: The more, the better. Preferably USB 3.0 ports.
-
Price
-
Power
-
GPU: Usually they are all low (good for GUI only).
B) Options
Now let's take a look at some available options:
Computer Name | OS | CPU | RAM | USB Ports | Price | Power | GPU | Thoughts |
---|---|---|---|---|---|---|---|---|
Raspberry Pi 3 Model B+ | Ubuntu Server, Ubuntu Core, Ubuntu MATE, Raspbian, ... | Quad-core Cortex-A53 1.4GHz 64-bit | 1 GB | 4 USB 2.0 Ports | $ 35 | 5V/2.5A | VideoCore IV | OS should be Ubuntu MATE or Raspbian (Not good as original Ubuntu). RAM is low for our purpose. |
Raspberry Pi 4 Model B | Ubuntu Desktop, Raspbian, ... | Quad-core Cortex-A72 1.5GHz 64-bit | 2 GB, 4 GB, and 8 GB | 2 USB 2.0 Ports, 2 USB 3.0 Ports | $ 45, $ 55, and $ 75 | 5V/3A | VideoCore VI | High RAM should be purchased. Good Option. |
Jetson Nano Developer Kit | Linux4Tegra (based on Ubuntu 18.04), Maybe Ubuntu 20.04 | Quad-core ARM A57 1.43GHz 64-bit | 4 GB | 4 USB 3.0 Ports | $ 100 | 10 W | 128-core NVIDIA Maxwell | Good option |
BeagleBone Black | Angstrom, Ubuntu, Debian | Single-core Cortex-A8 1GHz 32-bit | 512MB DDR3, 4GB 8-bit eMMC | mini-AB USB 2.0 client port, Type-A USB 2.0 host port | $ 50 | 5V/2A | PowerVR SGX530 | Old Tech/Not enough ports/Low CPU |
BeagleBone AI | Android, Linux, RTOS | Dual-core Cortex-A15 + 2 Dual Arm Cortex-M4 1.5 GHz 32-bit | 1GB RAM, 16GB eMMC | USB type-C for power, USB type-A host | $ 140 | 5V/2.5A | Dual-core PowerVR SGX544 3D | Not enough USB ports |
BeagleV | Fedora, Maybe Debian and Linux | RISC-V U74 Dual core 1.0GHz | 8GB LPDDR4 | 4 USB 3.0 Ports | $ 150 | 5V/3A | --- | Best BeagleBoard option. |
Google Coral Dev Board | Mendel Linux | quad Cortex-A53, Cortex-M4F | 4GB | Type-C OTG Type-A 3.0 host Micro-B serial console | $ 170 | 5V/2-3A | Integrated GC7000 Lite | A good option but limited to TensorFlow Lite |
Asus Tinker Board 2 | Debian 9, Android 10 | Dual 2.0 GHz Cortex-A72, Quad 1.5 GHz Cortex-A53 | Dual-CH LPDDR4 4GB | 3 USB 3.2 Type-A, 1 USB 3.2 Type-C | $ 180 | 15V/3A | Arm Mali-T860 | Powerful System! Professional |
C) Diving Deeper
The best options out of the aforementioned ones are:
-
Raspberry Pi 4 Model B
-
Jetson Nano Developer Kit
-
BeagleV
-
Google Coral Dev Board
-
Asus Tinker Board 2
There are several deeper important factors to be considered.
Here, several links are provided about these SBCs.
Jetson Nano having two parallel video decoders: Jetson Nano Specifications
Camera Developer (ArduCam) for RPi and Jetson: Jetson Cameras, Auto Focus Camera, Upcoming Camera, High Resolution