塑造未来的出行方式：NODAR 和安森美的自动驾驶愿景蓝图

五月 14, 2024

Safety, reliability and precision are paramount in the rapidly evolving landscape of autonomous driving (AD). onsemi and NODAR have collaborated to combine NODAR's advanced stereo vision technology with onsemi’s best-in-class imaging technology. To tackle pressing challenges core to the success and safety of driver-assisted and fully autonomous vehicles, onsemi’s Hyperlux™ image sensors now work with NODAR's development kit for those who want to get their hands on high-resolution stereo vision system. Together, we enable long-range, ultra-accurate object detection, such as the capability of finding small hazards, such as a rock, tire, or piece of lumber on the roadway, from distances of 150 meters and beyond, setting new standards for Level 3 AD capabilities at speeds up to 130 kph. In addition to exceptionally long-range 3D sensing, breakthroughs from NODAR and onsemi combine to enable vehicles to navigate more safely in adverse weather, during nighttime, and across unpaved or uneven terrains, significantly elevating road safety. Our combined efforts not only promise to advance vehicular performance but will also redefine the functionality of future autonomous vehicles, emphasizing seamless integration of technology.

What is unique and groundbreaking about each technology?

In this collaborative effort, we have integrated advanced technologies from both onsemi and NODAR to significantly enhance the capabilities of 3D sensing systems, setting new performance benchmarks for the automotive industry.

Sergey Velichko (onsemi): "At onsemi, innovation is at the core of our mission, particularly in ADAS and AD applications. Our Hyperlux technology delivers the industry's best low light performance and the widest high dynamic range (HDR), alongside superior image quality and stability across various automotive temperatures. Our joint solution includes two of our 8-megapixel Hyperlux sensors, which allow for longer range detection with high resolution and 150 dB HDR, doubling the detection distance compared to standard single camera systems. This capability is crucial for forward sensing applications and is equally effective at night, enhancing our ability to detect smaller objects and providing superior performance in challenging conditions."

Figure 1: Wide Stereo Vision System by NODAR and onsemi

Leaf Jiang (NODAR): "NODAR complements onsemi's technology with groundbreaking advancements in stereo vision. Our technology features frame-by-frame auto-calibration, optimizing the relative pose of two cameras over 22 dimensions and achieving precise, sub-pixel accuracy. Operating in real time and calibrating images every frame enables NODAR to maintain optimal performance across a variety of operating conditions, such as high vibration, inclement weather and low light. "

Long-Range Detection is Critical for Autonomous Driving

The advanced 3D detection enabled by the collaboration of onsemi and NODAR is crucial in high-speed environments, such as highways or on train tracks, where the ability to identify hazards from a distance is essential. Vehicles and trains traveling at high speeds require significant time and distance to come to a complete stop or to make a decision to avoid an obstacle. For instance, a train might need 1,000 meters or more to come to a complete stop if something is blocking the tracks. The joint solution provides the necessary range and lead time to detect and react to obstacles as far away as 1,000 meters. This capability ensures safer navigation, preventing accidents by allowing for timely and effective responses to potential hazards. Long-range object detection relies upon exceptionally high-resolution, crips imagery from cameras, and complex signal processing to align the images and extract depth information.

Sergey Velichko (onsemi): "The patented super exposure pixel technology of our Hyperlux sensors enhances color fidelity and image sharpness under all lighting conditions, enabling precise edge detection. The overflow technique used in these pixels extends the dynamic range, capturing true color in bright scenes while maintaining sensitivity in low light. The result is unmatched image quality, which includes LED flicker mitigation effective both day and night. Combined with NODAR’s high-resolution stereo vision, we achieve unprecedented precision in detecting object size and distance."

Leaf Jiang (NODAR): "Our enhancements to the Hyperlux sensors include a wide baseline stereo configuration and real-time calibration, which are crucial for maintaining high accuracy over long distances and under dynamic conditions. Our real-time calibration continually optimizes the stereo image pairs, a critical factor for accurate long-range detection."

Enhanced Depth Detection in Challenging Conditions

This collaboration between onsemi and NODAR offers a camera-based solution for depth sensing by using stereo image pairs to calculate the distance. This approach optimizes detection capabilities under adverse weather and ensures reliable performance across nearly all lighting conditions. Owing to the HDR and superior low light performance of onsemi Hyperlux sensors, NODAR’s stereo vision system can output better depth accuracy from stereoscopic imaging at a lower cost than other depth sensing technologies like LiDAR.

Sergey Velichko (onsemi): "Hyperlux technology is designed to excel in low visibility and inclement weather by reducing image sensor dark noise to sub-electron levels. This feature enables effective imaging in extremely low light conditions, such as on rural roads or when only using headlights at long distances. This capability is crucial for detecting small, potentially hazardous objects from afar. Additionally, the sensors maintain stability across a wide temperature range, from -40 °C to +105 °C ambient, ensuring dependable performance under various geographical and lighting conditions."

Leaf Jiang (NODAR): When applied to onsemi’s crisp images, NODAR’s patented auto-calibration algorithms are able to precisely row-align the images, enabling our proprietary stereo matcher to rapidly pick out common elements between the images and perform parallax calculations with higher precision than traditional techniques would yield. This in turn produces higher confidence depth measurements for more pixels than was previously possible and will dramatically improve the overall performance and safety of autonomous vehicles across a wider operational design domain (ODD).

The Future Landscape of Autonomous Driving Technology

As we look ahead, our focus is not simply on enhancing existing technologies, but also on exploring how we can support autonomy in the future across a variety of markets and applications.

Sergey Velichko (onsemi): "Our continuous innovation sets precedents in the automotive imaging industry. We're developing more advanced imaging solutions to tackle challenges in low light and inclement weather scenarios. In the near future, we anticipate introducing higher resolution sensors and more integrated features to enable even higher levels of autonomous driving, ensuring these advancements are both technologically advanced and financially viable."

Leaf Jiang (NODAR): "Expanding beyond traditional automotive applications, we are applying our stereo vision technology to sectors like industrial safety and agriculture. Our GuardView system utilizes this technology for 3D safety monitoring across diverse environments, providing high-resolution imaging and long-range coverage, which is essential for ensuring safety and enhancing operational efficiency in these fields."

To learn more about onsemi’s and NODAR’s latest technologies, join them at AutoSens & InCabin USA 2024: