Recommended Hardware

Commodity hardware to enable object detection and speed/direction measurement.

Customize the hardware to fit your needs! The core components include the computing device, storage, camera, and co-processor. Feel free to mix-and-match components, but most of the documentation and default configuration assumes using the hardware recommended below.

Sample Hardware Configurations

Here are some sample sensor configurations and the data it collects:

1. Camera + AI co-processor is the lowest cost and will give you object detection, direction, visual speed measurements, and much more.

2. Add in a radar for the most accurate speed and direction measurements and basic object detection for nighttime detection.

3. Include an environmental sensor to also measure air quality, gases, particulate matter, noise, temperature, and much more.

4. (future feature) Install only the radar for the most privacy-conscious built that will be capable of basic object detection, speed, and direction.

5. Add additional camera(s) to monitor more directions using the same AI co-processor.**

** The traffic monitor software is capable of supporting potentially any number of cameras either connected directly or via a local feed on the same AI co-processor for monitor multiple directions or any other configuration (see recommended hardware > cameras for more details). The TM software also has support for up to four (4) radars directly connected and paired in any pattern to the cameras.

Hardware Check List - Bill of Materials (BOM)

Use the following checklist as a quick guide to components you need to purchase

We are not affiliated with any of the stores or companies linked in this section. These are suggestions that have been used or tested by contributors. If you have used or tested more, post on TM GitHub discussions!

• Computing Device: Raspberry Pi 5 (RPi 5) 4GB/8GB

  • Power: official RPi 27W USB-C Power Supply

  • (Recommended) CPU Cooler: RPi5 active cooler

• Storage: microSD card (≥32GB),

  • Raspberry Pi SD Cards are modern, fast, stable microSD cards.

  • OR SanDisk Extreme Pro microSDXC UHS-I Card for larger sizes to store more videos and snapshots

• Camera: Raspberry Pi Camera Module 3

  • plus the RPi 5 Camera Cable

• AI co-processor: see AI Co-processor for options

• (Recommended) Radar: OmniPreSence OPS243-A Doppler Radar Sensor

• Enclosure: See Enclosure (weather-resistant box) for options

• (Optional) Air quality (AQ) sensor: Enviro+ with Particulate Matter (PM) Sensor

  • AQ sensor will need a male-to-female GPIO ribbon cable with the TM enclosure

• Screws: various M3 sizes for mounting board and enclosure, M2.5x10mm for Raspberry Pi mount

Computing Device

(Required) Raspberry Pi 5 (RPi 5) 4GB/8GB. The Traffic Monitor is designed around 4GB memory profile, but if you have many sensors and other applications running, 8GB may be more performant.

Also pick up a (very cheap) official CPU cooler: RPi5 active cooler which helps prevent overheating on very hot days.

The Traffic Monitor is based on the Raspberry Pi 5.

The Raspberry Pi 4B and earlier units are not recommended as they have experienced detrimental performance due to not meeting the power requirements on the peripherals (USB) for the TPU and radar for this setup. However, many have been successful with earlier versions of the Raspberry Pi for object detection, so your mileage may vary.

Storage

(Required) A high-quality microSD card or a SSD (see alternative). Recommend at least 32GB capacity for system files with minimal (or no) snapshot and video capture.

1. Option: Setup has been tested and works well with the SanDisk Extreme Pro microSDXC UHS-I Card.

2. Option: Raspberry Pi official SD Card performs particularly well but sizes only range up to 128GB.

3. Alternative: There are many options on the RPi5 to use a faster, more durable NVME (M.2) drive, including those that pair with the Coral TPU or other AI co-processors.

Power Supply

(Required) To run the Traffic Monitor and components.

The Raspberry Pi 5 is rated for 27-watts (5V at 5A) and using anything with a lower rating like the older RPi PSUs will result in resets and/or throttling. However, the Traffic Monitor typically consumes between 6-14-watts of energy when it is fully operational and inferencing. All components are intended to be powered off the Raspberry Pi 5, so a quality power source is recommended.

1. (Recommended) The official 27W USB-C Power Supply for testing and permanent mounts.

2. (Alternative) PoE (Power over Ethernet) HATs available for the RPi 5.

   1. Waveshare PoE HAT (F) or Waveshare PoE M.2 HAT+ has performed well for some contributors.

   2. You will need a PoE+ power supply, such as a PoE+ Power Injector.

3. (Optional) Battery for short-term off-grid--i.e. a few hours--we recommend a portable power station such as a Jackery and plug in the official 27W USB-C Power Supply.

   1. A 300-Watt-hours portlable power station should be able to last ~24-hours. (see calculation in custom battery below)

4. (Optional) Custom battery for longer-term off-grid use.

   1. Required components:

      1. Buck converter (step-down converter) to step down a 12v (or 24v) battery to 5v@5A for the Raspberry Pi 5

      2. LFP (LiFePO₄) battery at 12v or 24v with Amp-hours (Ah) to last desired time frame.

   2. Sample battery calculation (YMMV): the TM consumes ~13-watts, running it for 24-hour will require 13*24=312 Wh (Watt-hours), the LFP battery needs to be 312/12= 26, which would be 12-volts @ 30-Ah.

5. (Prototype) Solar panel + battery.

   1. Components: Same as custom battery above plus photovoltaic (PV) panel and solar charge controller

   2. Sample calculation:

      1. Find your daily Solar PV potential; e.g. Portland, OR has 7.15 and 1.42 kWh/day potential in an optimum summer and winter, respectively.

      2. Choose your PV panel size and calculate potential energy generation for the lowest time of year (winter in northern hemisphere).

         1. Potential * Size of PV panel = production per day; e.g. 1.42 * 0.1 (100-watt panel) = 142-Watts per day production in Portland, OR winter

      3. PWM solar controllers have ~80% efficiency on power conversion

         1. Production per day * controller efficiency = 142*0.8 = 113-Watts per day available.

      4. Calculate battery size requirements as above.

      5. Note: This is not enough to run the TM and charge a battery. You would require 3-times that amount, closer to a 300-watt PV in a Portland, OR winter.

   3. (Caveat) Solar production is very dependent on the amount of sun you are able to harvest. You need to consider factors like orientation/angle of panels, non-optimal days (cloudy, rainy), anything that blocks the sun such as trees, and more.

Camera(s)

(Required) For full object detection capabilities.

The official Raspberry Pi cameras are below recommended for low-cost, compact, local object detection; however any camera that can output H.264 is compatible with the traffic monitor, so you may attach USB or even networked cameras.

1. (Recommended) Raspberry Pi Camera Module 3 (wide angle recommended)

   1. Requires a RPi 5 Camera Cable that is sold separately.

2. (Alternative/additional) Raspberry Pi Global Shutter for faster motion capture and custom-lens based on your needs

   1. Requires a RPi 5 Camera Cable that is sold separately.

3. (Alternative/additional): See more at Frigate's recommended camera hardware.

The Raspberry Pi 5 has 2 (two) camera transceiver slots, so you can easily attach 2 native Raspberry Pi cameras.

See the Frigate camera setup for more information on tuning stream configurations based on various goals for your deployment.

AI Co-processor

AI Co-processor

(Required with camera) The AI co-processor is an efficient way to run the object detection model, much more efficient than CPU-alone.

The AI co-processor is used by Frigate to run the object detection model, see Frigate's supported hardware for more options and details. The Traffic Monitor assumes you are building on Raspberry Pi.

The below AI co-processors or detectors are capable of 100+ FPS with millisecond inference time, depending on the model and hardware. This means multiple camera feeds may be supported.

1. Recommended: Raspberry Pi AI HAT+ with Hailo-8 or Hailo-8L offers high-performance, power-efficient processing.

2. (Older alternative, becoming expensive, hard-to-find, and obsolete) Coral USB Accelerator is easy-to-use co-processor that you can connect to any computing device with a USB interface.

3. (Older alternative, becoming expensive, hard-to-find, and obsolete) Coral HATs (Hardware-Attached-on-Top [of a Raspberry Pi]) are more compact, upgradable than the USB accelerator:

   • Rapsberry Pi M.2 HAT+ pairs nicely with the Coral M.2 Accelerator B+M Key (not the A+E key!).

Radar

(Recommended) Provides accurate speed and direction measurement.

1. OmniPreSence OPS243-A Doppler Radar Sensor - certified with same tests as law enforcement speed radars. Detection up to 100-meters away.

Other Sensors

(Optional) For additional environmental data.

1. Air quality (AQ) sensor: Enviro+ paired with the (recommended) Particulate Matter (PM) Sensor. Also pick up a longer ribbon cable, we recommend the male-to-female GPIO ribbon cable.

Get AQ sensor details and capabilities on the Air Quality (AQ) Payload page.

Enclosure (weather-resistant box)

• Print it yourself: We offer a 3D printable model so you can build the truly open source Traffic Monitor. Visit our open source repository greendormer/tm-enclosure-3d for details and parts list.

  • 

• Purchase: (coming soon) Purchase the box or a kit to assemble yourself.

• Alternative DIY: There are many waterproof electrical junction boxes that may be modified to fit your needs with the traffic monitor. Rough dimensions to fit the Traffic Monitor components including the camera and radar should be around 9"x7"x4" such as the TICONN IP67 ABS Enclosure.