Thermal vs Day/Night ShootoutBy: Ethan Ace, Published on Nov 27, 2012
What should you do for dark outdoor scenes? Thermal cameras are getting a lot less expensive. Conventional cameras are getting better in low light. We decided to test these two categories - thermal and day/night - head to head to better understand what's better to use.
We tested in a large outdoor level field, overviewed below:
Three cameras were used:
- A DRS WatchMaster IP Elite, 320x240 imager, with a 40-degree field of view. Other cameras were matched to this field of view.
- Axis Q1602 with Computar F1.0 lens.
- Bosch NBN-932V with Fujinon F1.2 lens.
The DRS was selected because it is the least expensive thermal IP camera on the market while the Axis and Bosch were chosen do to their strong performances in previous low light shootouts (see the Q1602 test and the MP low light shootout).
The following scenarios were tested:
- Daytime, between ~4500-5000 lx (a partly cloudy day).
- ~0.1 lux with both visible cameras defaulted, 1/30s shutter speed, day/night set to auto.
- ~0.1 lux with the Q1602 set to 1/30s shutter speed, in black and white.
- ~0.1 lux with the Q1602 set to 1/2s shutter speed, in color.
- All tests were performed at ranges between 150-550', the specified limit of human detection range of the DRS camera.
With this approach, we could determine not only what was 'best' in low light but what tradeoffs existed during the rest of the day.
- In night testing, the least expensive thermal was still significantly better at detecting subjects than the leading SD camera when using the 'standard' 1/30s shutter.
- However, using a 1/2s slow shutter for the SD, the two cameras essentially tied, which was surprising given how extremely dark our night test conditions were. Tradeoffs still exist when using slow shutter and our examined at the conclusion of the report.
- While the 1080p HD camera was significantly better in the day than both thermal and SD, in the night test, at 1/30s shutter, no detection was possible even at the closest test point.
Conclusion: For long range, very dark conditions, a top low light SD D/N camera may be viable but only if used in slow shutter mode, recognizing the limitations of slow shutter. However, since a thermal camera only captures 'blobs' anyway, the comparable loss of a D/N with slow shutter is minimal.
Future Test: In December, we will do a shootout using add on IR illuminators to compare performance.
In the daytime, not surprisingly, even at 150', both SD and HD delivered significantly more details:
Thermal's relative performance improves at 300' as the color images get less detailed:
Increasing the distance to 450',
Finally, at 550', while the subject is hard to see with both the thermal and the SD, the 1080p camera shows the subject fairly clearly - a certain advantage for HD.
Low Light with 1/30s Shutter
After the sun went down, we tested the cameras using the same default settings: 1/30s shutter speed, with automatic switch to monochrome mode. This scene is approximately ~0.1 lux.
In all of these scenes, the 1080p camera did not deliver usable image even though the one used (Bosch) was the best performing 1080p low light camera we have tested so far.
At 150', the thermal camera shows the subject brightly and clear while the SD struggles to make out an outline.
At 300' the pattern continues:
At 450', it gets harder to make out the subject on thermal but still fairly obvious. By comparison, with the SD, it's nearly impossible.
Finally here's 550' with the trend continued:
Low Light with Color Mode
We also tried it with color mode on for the SD and black hot for thermal but results were roughly the same.
Here's a sample at 150':
And at 450':
Slow Shutter in Monochrome
Finally, we tried using a slow shutter to see the tradeoffs. Interestingly, a 1/2s slow shutter on the Axis Q1602 hardly delivered any improvements over a 1/30s shutter.
Here's at 150':
And at 300'
Slow Shutter in Color
Though slow shutter in monochrome mode provided little improvement, forcing the camera into color mode with a 1/2s exposure greatly brightened the scene, almost to where it felt like we were looking at a daytime image. However, our subject became extremely blurry when in motion, making our subject indistinguishable, even at close range. The improvement in brightness and this ghosting can be seen in this image:
For strict detection, however, this setting does much to bring the Q1602 on par with the DRS thermal. Objects are much easier to distinguish, as they're now shown as colors against the neutral background, instead of simply greyscale. This is evident at all ranges:
First At 150':
Then at 300'
Now at 450'
Finally at 550'
At 1/2s shutter, the image delivered of a standing (or slow moving person) is essentially a tie. However, clear tradeoffs / risks exist with putting an SD camera at such a slow shutter:
- Analytics will suffer numerous problems, likely forcing one to only manually / visually verify.
- Faster moving objects will likely appear as a blur. On the plus side, it should be easy to visually verify something is moving but it will be hard to see what it is.
Video Clips to Review
Watch the clips for yourself: