As cars become more intelligent, ensuring better vision has been the priority, especially at night. US-based Owl Autonomous Imaging integrates thermal images with artificial intelligence to do object detection, classification and ranging for a vehicle. The company launched its first commercial product in early January.
According to Wade Appelman, Owl AI's chief marketing officer, thermal cameras use passive energy – radiation from all objects in the form of heat – while RGB cameras use visible light.
He said thermal cameras can see in all light conditions, including in complete darkness. The sensors were created primarily for high-cost industrial applications like avionics and the military. But Owl AI has been developing cost-effective thermal cameras and AI software suitable for the automotive industry.
Appelman explained that real-time images shot by Owl AI's thermal cameras are run through the convolutional neural network the company has been training. The system will help a car detect an object, and then classify and check how far it is away. The information is essential for vehicles to react safely to ever-changing road situations.
The first version of Owl AI's solution can classify cars, children and adult-sized people, evaluating their distance from a vehicle precisely, Appelman said. The startup will continue to enhance the technology to classify deer, stop signs, traffic cones and many more objects.
Detecting unexpected objects at night has been a real problem for drivers. There is also room for improvement in the related car safety features. The US Insurance Institute for Highway Safety did a nighttime test of pedestrian automatic emergency braking (AEB) systems in 2022. Results showed eight of the 23 tested models earned basic scores. In addition, four vehicles did not perform well enough in the dark to get any score.
Appelman said vehicles that failed to stop in time in the test were mostly equipped with RGB cameras or radars.
Singling out objects of interest to enable more computing power
Owl AI was founded in late 2018 by experts in thermal imaging and computer vision software. Backed by US insurance company State Farm Insurance, it attracted great interest from automotive OEMs and Tier-1 suppliers at CES 2023. Appelman said the company has been building its thermal imaging chip and software in past years.
As for price range, he expected automotive thermal cameras to be slightly more expensive than radars and RGB cameras but much cheaper than LiDARS.
Owl AI's solution can single out objects of interest in classification and ranging data in just one thermal image stream, Appelman said. Other solutions usually need to fuse different streams, with a radar providing the distance information and a camera for classification, for example.
The CMO added that singling out objects of interest allows the system's microprocessors to deal with only the essential information and save computing power. Therefore, more spare cycles will be available for other functions. Less power consumption is required and less waste heat will arise.
Targeting US, European, South Korean and Japanese markets
At CES 2023, Owl AI released its first commercial product available to the general public. It was an evaluation kit for the company's Thermal Ranger ADAS & Autonomous Navigation Development Platform.
According to an announcement, the hardware and software kit enables automotive OEMs and Tier-1 suppliers to evaluate Owl AI's solution for their uses in Pedestrian Automatic Emergency Braking (PAEB) and other ADAS features supporting up to Level 4 autonomous driving.
Appelman said the company is now offering a thermal camera with VGA resolution. A high-definition camera will be coming later this year.
Owl Al targets the markets of Europe, South Korea and Japan besides its home turf, the US, according to Appelman.
"It's the automobile manufacturers that are looking to add safety sensors to their cars in order to see things better at night than they can with today's sensors," he said.
The US-based startup also looks forward to working with lens manufacturers in Taiwan to scale up. Appelman said Taiwan hosts some of the best lens companies which produce in high volume with low costs. Working with these companies is critical to Owl AI's strategy to meet its customer needs.
He also said the company now has two foundry partners and does packaging and testing in the region that includes Taiwan. As it moves to mass production for automobiles, Taiwan will be an ideal candidate for assembly and testing. Moreover, he said TSMC is a great potential foundry for manufacturing OW AI's readout IC.
Wade Appelman is Owl AI's chief marketing officer. Credit: DIGITIMES