For intelligent traffic systems, VGA resolution CCD-based cameras have traditionally been the accepted technology, but with the quality of CMOS image sensors continuously improving, CMOS-based cameras are now a viable option for demanding license plate recognition applications. The latest system requirements are for higher resolution (HDTV) and frame rates of 30 to 60 frames per second as well as all of the following: color, NIR sensitivity, high dynamic range, and low noise at high gain. Imaging systems for traffic are exposed to extreme lighting and temperature conditions which affect the image quality. So what are the advantages and disadvantages of choosing a camera based on a CMOS image sensor versus a CCD image sensor?
An overview of our comparison of CCD vs. CMOS:
|Advantages of CCD||Advantages of CMOS|
Advantages and Limitations of CCD-based cameras
For many years, CCD-based cameras were the only available option and are therefore an accepted and familiar technology. CCD continues to offer several advantages over CMOS. CCD image sensors offer a perfect global shutter which is critical for freezing fast movements. Also, CCDs have high Qe and low read noise (good sensitivity in NIR) which is beneficial in low light situations and with NIR illumination which is commonly used. Dark current increases at higher temperatures and can add noise to the image. CCD has lower dark current than CMOS.
With images of cars on the road, there can be very bright objects in the image such as the sun or headlights, and then very dark portions of the image in the license plate. This requires a wide dynamic range for handling huge contrast within one image. CCD has good dynamic range performance, but because CMOS has each pixel circuitry, greater influence can be imposed to have even higher dynamic range. Also, the bright spots in the scene can create vertical lines in the image and obstruct the viewing of the license plate because of blooming and smear with CCD technology.
Advantages and Limitations of CMOS-based cameras
CMOS has no blooming and smear as well as very good dynamic range within images as mentioned above. Other advantages for CMOS include significantly higher frame rates and lower power. There have been improvements in CMOS to allow for better sensitivity in the NIR.
There are now very good CMOS sensors with a global shutter rather than a rolling shutter, but the global shutter is still being perfected. Even though there is no blooming and smear with CMOS, CMOS has other image artifacts. The total noise level of a CMOS sensor is typically greater than for a CCD image sensor. Fixed Pattern Noise (FPN) and the other noise increases with higher temperatures and the noise is affected so that if the CMOS sensor has double noise than CCD, the effect is multiplied. Again these image artifacts are reduced in subsequent generations of sensors.
In 1 to 2 years, CMOS image sensors will outperform CCD regarding traffic requirements in high speed applications. It may then take some time for CMOS to take over in systems as the market accepts the change and makes all of the corresponding system changes.
As CMOS improves, it will be even more valuable in non high speed applications because of the interscene dynamic range and lower power. Lower noise will be available in the next few years.
CMOSIS with AR coating would be especially interesting as it would avoid ghost images. Unlike in the CCD vs. CMOS in the machine vision and military markets, we see an even greater portion of the traffic market moving to CMOS and more quickly.
A related article: