ICs & Semiconductors: 4G CMOS imagers enrich driver experience, safety.

The first three generations of CMOS imagers were focused on improving image quality to match CCD levels of performance. Some CMOS imager vendors have already attained this level. Today, the goal is not only to improve the photoplane’s performance but also to find ways to better fit the imager into emerging applications such as automotive electronics.



In the next decade, it is projected that cars will have four to 16 cameras in them. (Figure 1 shows some of the camera applications of future vehicles.) One problem, though, is that customers are looking for military-qualified components at toy prices. This is unachievable unless you can come up with ways to reduce the overall system cost, which is exactly what is being addressed today.SCENE VIEWING

Currently there are two types of applications for cameras in vehicles – scene viewing and scene understanding. Scene viewing applications are those that capture an image and present it to a screen inside the vehicle to help the driver make decisions based on the environment around him. These applications include rear-vision assistance, mirror replacement, blind spot, side view and post-accident status cameras. The imagers for these applications are focused on presenting accurate, high-quality images that the driver can easily understand.



Initially, a wide-angle view lens CCD camera is mounted on the trunk of the vehicle. This camera passes NTSC/PAL-based video to an electronic control unit (ECU). The ECU uses the ADC to digitize the analog video and then puts the image into a processor to correct lens distortion. After that, the processed video stream is put back into the analog domain and converted back into NTSC or PAL to be sent to the display in front of the cabin.



REAR VISION

Micron’s 4G CMOS imager is specifically designed for automotive scene viewing applications. It operates from -40°C to +105°C, and delivers NTSC/PAL video as well as digital formatted video. Also, the chip lets the user decide if he wants the video coming from the photoplane or from the digital video input port to use the on-chip DAC and NTSC/PAL formatter (Figure 2).



This architecture enables the system designer to take the digital video directly from the imager and feed it into the processing chip that will correct the lens distortion and add overlays. The post-processed video is then fed back into the imager and uses the on-chip DAC and NTSC/PAL encoder to send the now post-processed NTSC/ PAL-formatted data to the front of the vehicle. By doing this, the connectors of the ECU, ADC, DAC and NTSC encoder are not needed. Hence, system cost is reduced on the order of $5 to $10 without increasing the cost of the imager.



SCENE UNDERSTANDING

The other set of applications for vehicles is based on scene understanding. In many of these applications, the driver will never see the imagery because the video is sent directly to a processor. The processor analyzes the features of the scene and makes decisions based on them. These decisions are then fed to various vehicle control devices which control lane tracking, urban cruise control, collision warning, avoidance, rain sensing, headlamp dimming, drowsy driver, driver attention, occupant positioning for airbag deployment, gaze detection, mirror adjustment based on where your eyes are, and biometric ID. In many cases, decisions are being made faster than a person could react, or are being made as the driver is in crisis situation and needs assistance. To do this, the CMOS imager must have a wide dynamic range to be able to see lane markers while driving with the sun low on the horizon. Also, to present the image with as little distortion as possible, a global shutter design is needed. Figure 3 shows the distortion effect of a rolling shutter vs. global shutter when capturing a video of a spinning cube. With processing you can make the image on the right look like the image on the left, but since Micron’s 4G CMOS imager is designed with the vehicle application in mind, that entire processing burden has been resolved. This enables the system designer to do more processing to increase accuracy, add another application, or choose a less expensive processor. The US National Transportation and Safety Board estimates that more than 70 percent of rear-end collisions could be avoided if the vehicle started braking 0.5 seconds sooner. Since it would be difficult to make faster people, the vehicle should decide when to brake to avoid collision. Aside from safety reasons, imaging-based systems for vehicle applications also make the trip more pleasurable and stress-free.



For these applications to be widely adopted, imagers need to be designed using a system level approach. It is just a matter of time before 4G CMOS imagers play an integral role in our driving lives.



Click here for Figures:



Figure 1, Figure 2, Figure 3