Lecture

2.3 ADAS and Autonomous Driving: Comparative Analysis of Super Exposure and Split Pixel Automotive 8 MP HDR LFM Image Sensors for Object Detection

  • at -
  • Future Control Stage (C5.259)Next Mobility
  • Language: English
  • Type: Lecture

Lecture description

Evolving regulations including Euro NCAP and NHTSA are directing the automotive industry to sharply adopt and transition to higher levels of safety and autonomy reducing fatalities and injuries on roads. The latest car OEM implementations of ADAS and autonomous driving systems (ADS) rely on an array of automotive 8 MP cameras that form a sensing cocoon for various object detection and recognition. These objects include vulnerable road users (VRUs), other vehicles, road debris, curbs and traffic lights, signs, and lane markings. Maximum vehicle speeds define minimum sensing distances at which these objects should be detected, from slow urban street speeds to 140 kph on a highway. In this study we focus on two major 2.1 µm pixel architectures for latest 8.3 MP high dynamic range (HDR) with LED/light flicker mitigation (LFM) image sensors: super exposure pixel and split pixel. By examining both architectures, we extract their major technical parameters and discuss the differences across typical automotive use cases. Furthermore, we apply well recognized ideal observer SNRI metric, to study detection abilities in typical automotive scenarios for VRUs and small objects. Using cameras with different field-of-views we demonstrate that latest Hyperlux 2.1 µm single optical stack super exposure pixel sensors outperform other solutions and enable detection of these various objects at longer distances thus enabling up to 140 kph speeds and level 3/3+ of autonomous driving. We compliment the studies with numerous images and videos. With these case studies, the audience is provided a deeper dive into onsemi’s new generation of automotive Hyperlux 2.1 µm sensors that support automotive systems with 150 dB HDR LFM, low light performance exceeding human eye, and high-fidelity color images. Audience will get to view how Hyperlux sensors address corner cases by leveraging higher probabilities of detection in most demanding use cases.

Speaker

Further lectures