License Plate Capture Shootout 2014

What should you be using to capture license plates consistently?

We tested 3 major types of cameras head to head to see the tradeoffs:

• Super Low Light cameras: Day / night MP cameras with advanced night time performance have emerged over the past couple of years. We selected the Samsung WiseNet III 5004 camera as the representative here.
• Integrated IR cameras: MP cameras with integrated IR illuminators are increasingly common. We selected the Avigilon Integrated IR offering here.
• Purpose Built License Plate cameras: These are the least commonly available among IP cameras but are designed to capture license plates out of the box. We selected the Geovision MP LPC camera here.

We tested:

• Daytime
• Night time
• Low speed (< 10mph)
• High speed (>40mph) 

Those tests let us understand the tradeoffs during different times of the day and conditions.

Here is a sample of one of the 12 image comparisons inside:

For background on fundamental issues in capturing license plates, see:

• License Plate Capture Fundamentals Tested
• How to Handle Headlight Surveillance Problems

Key Findings

Here are our key findings from this test:

• During the day, all cameras were able to reliably capture plates at speeds up to 40-45 mph (the maximum we tested), using default settings. 
• Out of the box, the GeoVision 1.3 megapixel license plate capture camera reliably captured plates at both at slow speeds, >10 mph, and faster, 40+ mph, with no misses front or rear, day or night across our two lane wide field of view, about 22'.
• Using default settings, the Avigilon H3 1080p bullet camera failed to capture plates at either slow or fast speeds at night, though daytime capture was reliable.
• Speeding exposure to 1/500s, the Avigilon bullet reliably captured both front and rear number plates at low and high speeds. However, this greatly reduced the brightness of the image at night, making it less useful for general surveillance of the scene.
• The low light Samsung SNB-5004 was unable to capture plates at night, even when exposure was sped to 1/1000s. Headlights, taillights, and even small license plate lights caused bloom and overexposure at the camera's high gain settings. Reducing gain to low (default is "high") improved these issues, but capture was still unreliable, and usefulness for surveillance of the overall scene reduced.
• The SNB-5004 provided vehicle color information at night, one advantage which no other cameras had, since the Avigilon bullet is day/night and GeoVision GV-Hybrid 10R is monochrome only.

Pricing

Camera pricing compares as follows:

• Avigilon 2.0W-H3-B01: ~$650 USD estimated street price
• Geovision GV-Hybrid 10R: ~$850 USD online
• Samsung SNB-5004 w/Tamron 5-50mm M13VG550 lens [link no longer available]: ~$560 USD ($380 + $180 online)
  

Recommendations

Based on our tests, at ~40 mph and below integrated IR cameras may be the best option for license plate capture due to their lower cost and greater flexibility, providing better scene awareness than dedicated capture cameras. However, there are two caveats to using bullet cameras for capture:

• Bullet cameras will likely require adjustment of shutter speed for reliable capture, with 1/500s being the minimum we found in our tests.
• The camera used in our test, the Avigilon 2.0W-H3-B01 features smart IR, adjusting power based on objects in the scene. Cameras without this feature may require further speeding of the shutter to prevent IR overexposure, or not capture at all.

Specialized license plate capture cameras may be the best option at higher speeds (>40 mph), though this is most typically seen in traffic surveillance, of highways or public thoroughfares. These cameras' fast exposure and resulting darker imaging compared to integrated IR cameras make them unsuitable for general surveillance of the scene, as well as identifying vehicle make and model.

Super low light cameras are not recommended for license plate capture. They may work, with greatly reduced gain and fast shutter settings, but these settings essentially eliminate the benefits of their increased low light performance. However, these cameras may provide the best overview and vehicle information of these three options, due to their color performance in low light.

High Speed Performance

We began in an outdoor scene, testing with vehicles moving 40-50 mph, across two lanes of traffic, about 22' wide, testing front and rear plate capture performance.

In this scene, all cameras are able to capture rear plates without issue.

The front, European-style plate was no issue, either. 

 

Nighttime Comparisons

At night, with lux levels substantially darkened to >1 lux, we tested in the same scene. 

Using default settings, only the GV-Hybrid was able to capture the plate, with blur and overexposure obscuring plates on the other two cameras due to their much slower exposure (1/30s vs. 1/500s).

 

Speeding exposure to 1/500s in the Avigilon and Samsung cameras, the Avigilon H3 bullet is able to capture license plates, though the Samsung SNB-5004 is still extremely overexposed. 

The front plate was also easily visible to the Avigilon bullet at 1/500s shutter, though the SNB-5004's image was further overexposed by headlights. The added resolution of the Avigilon bullet (1080p vs. 720p) helps it here, with characters slightly more legible than the Geovision GV-Hybrid.

The images below show our attempt to configure settings on the SNB-5004 so it may capture plates. Even at 1/1000s, blooming of vehicle lights obscures the license plate. Manually reducing gain from high to low provides marginal capture, but eliminates the performance advantages these cameras have in low light.

Slow Speed Comparisons (>10 mph)

We tested the cameras' performance at slower speeds, 5-10 miles per hour, typical in parking scenarios, again across approximately two lanes, a 22' horizontal FOV.

Daytime

All cameras are able to capture plates in this scene without issue, both rear:

As well as front:

Night

We tested in this same FOV at night, as well:

Cameras other than the GV-Hybrid 10R were once again unable to capture plates using default settings, seen in the comparison below.

We sped the shutter to 1/500s as we did in our higher speed tests. At this shutter speed, Avigilon is easily able to capture the rear plate, while Samsung remains unable.

The same was true of the front plate, as well, with headlights completely overwhelming the SNB-5004.

Shutter Speed Scene Impact

As mentioned earlier in the report, raising shutter speeds to 1/500s affects the overall brightness of the scene, making the entire scene darker. For example, we have compared the Avigilon's default shutter speed against 1/500s shutter speed.

We also took a look at the Samsung SNB-5004's image when changing from a default shutter speed up to 1/500s.

Video Examples

The following are examples of nighttime plate capture with all cameras at varying settings.

First, the Geovision GV-Hybrid 10R, successfully capturing using default settings, though little else can be seen in the scene:

Avigilon 2.0W-H3-B02 using default settings, unable to capture:

Avigilon with maximum shutter speed set to 1/500s, reliably capturing:

Samsung SNB-5004 using default settings, completely unable to capture:

Samsung 1/1000s with low gain, marginally able to read some characters, but not reliably:

Readers may also download a set of these clips, as well as daytime clips, in this 60 MB .zip file. Note that these clips will work on Windows only, as they use the Exacq embedded viewer.

Methodology

Cameras were tested using default settings, unless otherwise noted. Avigilon and Samsung cameras started at 1/30s shutter speed, which was sped where necessary. Geovision shutter speed was left to automatic, no slower than 1/500s. Cameras were tested one at a time so IR illumination interference was not an issue, with the vehicle moving past at the specified speed multiple times to check capture.

Cameras were integrated to Exacq version 5.10.4.55064 without issue. The following firmware versions were used:

• Avigilon 2.0W-H3-B01: 2.6.0.14
• Geovision GV-Hybrid 10R: 1.02 2013-08-26
• Samsung SNB-5004: 1.13_131218

Quiz

Take this five question LPC quiz to reinforce your understanding of the key findings and results of this report.