Ambarella’s low-power and high-resolution video compression, image processing, and deep-neural-network processors and software enable cameras to become more intelligent by extracting valuable data from high-resolution video streams.
By Murray Slovick, Contributing Editor
Ambarella Inc., a developer of high-resolution video processing and computer vision semiconductors, kicked it up a notch with introduction of the CV22AQ automotive camera system-on-chip (SoC). Target applications for the part, which features the Ambarella CVflow computer-vision architecture for deep-neural network (DNN) processing, include front ADAS cameras, electronic mirrors with blind spot detection (BSD), interior driver and cabin monitoring cameras, and around view monitors (AVMs) with parking assist.
According to the company, the new SoC provides the performance necessary to exceed New Car Assessment Program (NCAP) requirements for lane keeping, automatic emergency braking (AEB), intelligent headlight control, and speed-assistance functions. Fabricated in 10-nm process technology it supports the small form factor and thermal requirements of windshield-mounted forward ADAS cameras.
The CV22AQ’s CVflow architecture provides computer-vision processing in 8-Mpixel resolution at 30 frames/s to enable object recognition over long distances and with high accuracy. CV22AQ supports multiple image sensor inputs for multi-FOV (field of view) cameras and can also create multiple digital FOVs using a single high-resolution image sensor to reduce system cost. It enables DNNs for object detection, classification (i.e., of pedestrians, vehicles, traffic signs, and traffic lights), and tracking, as well as semantic segmentation for applications such as free space detection.
The CV22AQ boasts a quad-core 756-MHz Arm Cortex-A53 processor and a multi-channel image signal processor (ISP) with up to a 576-Mpixel/s input pixel rate that facilitates imaging in low-light conditions. High dynamic range (HDR) processing extracts maximum image detail in high-contrast scenes.
The SoC includes 4K/8-Mpixel encoding in both AVC and HEVC video formats, allowing customers to add video recording and streaming capabilities to their automotive cameras. Its cybersecurity features, which include secure boot, TrustZone, and I/O virtualization, enable over-the-air (OTA) updates and protect against hacking.
“To date, front ADAS cameras have been performance-constrained due to power consumption limits inherent in the form factor,” said Fermi Wang, CEO of Ambarella. “CV22AQ provides an industry-leading combination of outstanding neural-network performance and very low typical power consumption of below 2.5 W.”
Interfaces include Gigabit Ethernet, USB 2.0 host and device, dual SD card controllers with SDXC support, HDMI v2.0, and MIPI DSI/CSI-2 4-lane output. The set of tools provided to help customers’ port their own neural networks onto the CV22AQ SoC includes a compiler, debugger, and support for standard machine-learning frameworks such as Caffe and TensorFlow, with extensive guidelines for DNN performance optimization.
CV22AQ meets the AEC-Q100 grade 2 (−40 to +105°C operating temperature) automotive standard and is currently sampling to tier-1 customers and tier-2 algorithm providers. Chip samples with ASIL-B support are targeted for availability this year.
Partnership Formed with HELLA Aglaia
In addition to launching the CV22AQ automotive camera SoC, Ambarella has joined forces with intelligent visual-perception software developer HELLA Aglaia to deliver a complete computer-vision software stack. The stack will leverage a platform based on CV22AQ technology that will add artificial-intelligence (AI) features in front ADAS cameras.
HELLA Aglaia is a full subsidiary of HELLA, a German tier-1 supplier based in Lippstadt, Germany. HELLA previously announced partnerships with NXP and Renesas in the field of ADAS car vision platforms. HELLA Aglaia’s ADAS platforms, which include NXP’s S32 and i.MX auto-grade processors, expects the next step in the cooperation will be to add AI for automated driving.
Similarly, HELLA’S open and scalable front camera solution for ADAS and automated driving with Renesas Electronics Corp. combines the Renesas R-Car V3M, high-performance, low-power image recognition SoC and HELLA Aglaia´s camera software designed to meet level 2 (partial automation) and level 3 (conditional automation) of the SAE International’s J3016 standard.