Imaging moving plates in low light is the most challenging application in the CCTV industry. It’s one thing to use expensive cameras and IR to snag fast moving plates on the freeway, but getting the same detail on the street without the support of visible light defeats all comers.
WE test quite a few cameras at SEN, usually top end devices. A weakness of our application is that it never changes. But it’s a strength, too, in that it allows us to say, with many years experience, that getting moving plates on the street in 4-10 lux is too much of a challenge for an unassisted standard CCTV camera. You can tweak whatever you like on a standard camera but it’s unlikely to help.
It’s easy to say that the issue is down to slow shutter speeds causing motion blur but there’s more going on than just shutter speed. Things like slower zoomed apertures, the use of slimmer H.264 and H.265 compressions instead of MJPEG, which can do better with motion, the presence of amplification noise, poor management of tone mapping, sensors that are prone to blooming with fast apertures and plenty more.
Something worth pointing out is that engineering teams build CCTV camera engines for a mid-point – many installers put them in at near default – even live control rooms aren’t always tweaking camera settings in real time. Mid-point camera settings simply aren’t extreme enough for moving plates. For that you need big sensors, specialised compression protocols; longer, faster fixed lenses – and something that’s probably not considered often enough, the combination of capable cameras carefully set up, supported by visible light and traffic calming strategies that slow vehicle speed down to 5-15kmph. In some applications where plates are the key, the management team needs to invest in dedicated camera technologies.
Moving faces aren’t easy in low light, either.
According to Oshana Jouna of Axis with the right camera, settings can make a large difference when it comes to license plate capture.
“Specialized license plate cameras are delivered with suitable default settings and require a minimum of tuning,” Jouna explains. For other cameras, the following settings might need to be changed including the following:
1. Max shutter time:
Vehicles which are moving through the image will cause motion blur, as below, if the shutter time of the camera is too long. The maximum shutter time depends on the alignment of the camera as well as the speed of the vehicles.
A car moving at high speed is imaged with an exposure time of 1/30 s – the recommended max shutter time depending on the angle between the camera and the car travel direction, and on the speed of the vehicles. Recommended max shutter time depending on camera angle and car speed. 1 ms = 1/1000s.
2. Max gain:
Since the license plate is made of reflective material it will shine up brightly when exposed to intense IR light. The surroundings will be much darker since other objects reflect much less light. The result can be that the license plate gets overexposed and is impossible to read.
The max gain setting will determine how the license plate is exposed at night. Exactly how to set max gain depends on the available IR intensity, the distance to the vehicles, and the light sensitivity of the camera. Somewhere between 9 dB and 21 dB gives reasonable results when using the built-in IR of an Axis camera.
3. WDR:
Wide dynamic range (WDR) comprises different techniques to increase the dynamic range of an image. WDR is very useful to bring out details which would otherwise hide in shadows, or to prevent the camera getting blinded by strong light.
WDR can cause motion artefacts in images of moving vehicles, depending on how WDR is implemented in a specific camera. If the camera specification does not say otherwise, we recommend that you always switch WDR off for license plate capture.
In summary, fine tuning a camera for LPR applications can be a time consuming and overwhelming task. Axis cameras that are suitable for LPR come with AXIS Traffic Wizard. AXIS Traffic Wizard is an application that optimizes the image settings based on the positioning of the camera and the light conditions at the scene. These settings will then be used in the Traffic overview Scene profile.
According to Andrew Cho of EOS Australia, the best camera settings for plates on the basis of using 1/2.8-inch sensor at a location with no lighting include a minimum shutter speed of 1/1500sec with AGC low, and IR lighting set to its maximum performance.
“Best to set medium level for the noise reduction,” Cho says.
This is actually super performance from Hik, with the camera at right angles to a vehicle moving at 30kmph – IR is activated here.
Meanwhile, Ryan Shi of Hikvision says that if he had to select the perfect camera settings for a low light application – where face ID unassisted was also required – there would be a number of considerations.
“For facial and number plate application minimum pixels need to be ensured for analysis,” Shi explains. “So, vari-focal lens, minimum 2MP resolution, 1/1.8inch sensor and aperture of F1.4, or larger, would be important to achieve the goal. Moreover, on-board analysis also helps improve the efficiency and cost.”
“Fast shutter speeds are very important to freeze the image for number plates. It would be much better to choose a low-light camera to ensure the image both during the day and the night. Appropriate focal length and installation are also important to ensure number plate recognition.”
Meanwhile, Hills’ Anitha Bellary says that camera performance is important.
“A ½-inch image sensor is a must,” Bellary says. “For LPR, strong IR and AE control are also a must.”
What are the best camera settings for moving plates, in Bellary’s opinion?
“Franking speaking, it is highly dependent on the LPR engine you use,” she says. “LPR software is designed to calibrate the camera settings that work best for LPR automatically – this should be done by software automatically, not by user settings.”
Conclusion
It goes without saying that there’s nothing easy about capturing moving plates on the street. Our experience with standard CCTV cameras is that head-on angles, directional light and slower speeds are the trick. For instance, the only non LPR CCTV camera that has ever delivered reliable plates in a SEN test in our sub-2 lux back lane application is Sony’s SNC-VB642D, a bullet camera that’s assisted by motion-activated white light. We think the Bosch MIC7000i supported by its white light array would do the same, but in our test a couple of years ago, the camera location wasn’t ideal for that application.
#sen.news
