Image Sensors: The Eyes For Autonomous Vehicles

The IEEE Women in Engineering International Leadership Conference (WIE ILC) was filled with a slew of informative sessions grouped into four themed tracks: Executive Leadership, Empowerment, Innovation, Disruptive Technology, and Workshop.

Radhika Arora, Sr. Manager at ON Semiconductor—Image Sensor Group (ISG), led a talk that fell into the Disruptive Technology category titled, “Image Sensors—Eyes for Autonomous Vehicles.”

According to the IEEE WIE ILC description, the presentation “investigates the challenges and opportunities pertaining to imaging that may arise as a result of emerging autonomous vehicle (AV) technologies.” AV technology will bring fundamental change, create disruption, and have far-reaching impacts. Where do image sensors fall into the equation to make the AV segment successful?

Arora began by saying over the last five years, the talk of autonomous tech has evolved. Initial chatter started as science fiction, but now self-driving cars have shifted to an inevitability.  

“In California, 54 companies have been grated permits to test autonomous technologies on the streets. These companies are redefining modern mobility,” says Arora.

Voyage was chosen as one of the examples, which has a self-driving taxi service at The Villages, Florida. Waymo followed, which has a public trial of self-driving vehicles under its Early Rider program in Arizona. Ford and Domino rounded out the examples, with their human-free delivery service.

With all these trials, what is the potential impact of autonomous driving? “AV technology will have an estimated seven trillion dollar impact on the global economy,” Arora says.

According to Arora, autonomous cars must perform better than a human. It needs to have a 360 view without any blind spots. To achieve this, Arora points out three essential sensors: image, lidar, and radar.

The talk honed in on image sensors, which are advantageous since they can distinguish color. This proves useful analyzing traffic lights and lane markings. According to Arora, image sensors are attractive to the self-driving market because of its low cost. However, image sensors have trouble in bad weather and low-light conditions.

The trends of autonomous vehicle cameras were outlined as the following: More Cameras, Wider Field of View, More Pixels, More Advanced. “The bottom line is, the more pixels you have the smaller the object you can detect at a further distance,” Arora adds.

The talk ended with cybersecurity of an image sensor. “Hacking image sensors basically means you’re blinding the car. A denial of service can make the car go blind,” says Arora.

She outlined the following attacks during the presentation:

1. Denial of Service: Loss of video, no signal, black/white screen.
2. Tampering: Changing video data to add/remove/modify objects in view.
3. Repudiation: Cause receiver to reject or discard data.
4. Eavesdropping/Privacy: Capturing video content for malicious purposes (tracking, etc.).
5. Spoofing: Fool receiver into accepting data from an untrusted source.

Although promising, there is a broad spectrum of issues from safety to environmental conditions that self-driving car sensors need to address before we permanently take our hands off the steering wheel.