Dual-camera system monitors driver and road with single AI SoC - Embedded.com

Dual-camera system monitors driver and road with single AI SoC

KeepTruckin’s new dashcam, based on Ambarella’s CV22 system on a chip, uses AI to monitor drivers of trucks and other commercial vehicles.

Driver monitoring systems have become an essential part of the conversation in automotive safety. While much of the driver-monitoring discussion has been around its use for individuals in passenger cars, there’s a huge market in the commercial vehicle sector for monitoring drivers in their trucks to ensure safety and try to prevent accidents, as well as to enable fleet managers to assess risk and drive down insurance costs.

Serving this market, fleet management technology company KeepTruckin this week said it partnered with edge artificial intelligence (AI) vision system on chip (SoC) firm Ambarella to develop its new aftermarket dashcam for larger commercial vehicles. The new dashcam uses a single Ambarella CV22 SoC to simultaneously provide AI and image processing for its dual-camera system, which integrates one camera for the front advanced driver assistance system (ADAS) with incident recording, and a second RGB-IR camera for the driver-monitoring system (DMS) with driver recording.

The CV22’s performance-per-watt specification also enables the small AI dashcam to run KeepTruckin’s proprietary AI algorithms for real-time high-risk behavior detection and active warnings directly on device with minimal heat dissipation.

EE Times spoke to executives from both KeepTruckin and Ambarella to find out more about the product development. Jai Ranganathan, SVP for product at KeepTruckin, said, “Driver safety has become more important than ever, especially as there’s more litigation against commercial operators. So, the concept of dashcams has become very popular. We partnered with Ambarella to develop our next generation dashcam. A key metric for us when looking at their chip is inference per Watt; performance relative to cost was also important. We needed to do things like depth detection with a single lens, which can be difficult.”

He said that the whole process, from inception to having the dashcam out today, took around 18 months. “The biggest factor for us [in enabling this quick time to market] was having a partner that could help with the whole development process. It wouldn’t have been feasible to get to market as quickly otherwise.”

KeepTruckin’s director of product, Abhishek Gupta, added, “The developer support was extremely beneficial, especially since we built this during a pandemic when many people were working remotely.”

Ranganathan said KeepTruckin’s AI dashcam leverages Ambarella’s edge AI and vision processing performance to deliver advanced AI-based features that detect high-risk behavior with high accuracy and efficiency. “Our continued fleet management innovation is enabled by Ambarella’s scalable range of CVflow AI vision SoCs, which are all supported by a common, robust SDK, enabling a platform approach to our product development.”

Providing the analyst perspective, Ian Riches, VP for the automotive practice and director at Strategy Analytics, commented, “KeepTruckin’s selection of Ambarella SoCs for its new AI Dashcam gives it a powerful, cost-effective and efficient platform on which to build. Ambarella’s ecosystem gives it the tools required to develop tailored AI solutions that will be of benefit to its fleet customers.”

Front ADAS features enabled by the CV22 include warnings for close following, lane departures, forward collisions, speeding and traffic violations. Using the same SoC, running multiple simultaneous neural network models, the in-cabin camera’s DMS capabilities include monitoring for driver fatigue, distraction and policy violations, such as contextual cell phone use or seatbelt monitoring, in combination with data from the front camera.

The AI Dashcam is connected to the KeepTruckin Vehicle Gateway, which uploads the pre-analyzed data, video and still images to KeepTruckin’s cloud-based fleet management software in real-time. The CV22 SoC integrates Ambarella’s image signal processor, which provides high quality, 1440p resolution HDR videos across all lighting conditions, while utilizing its on-chip H.264/H.265 encoding to reduce transmission bandwidth and storage costs.

KeepTruckin can upload additional features to the CV22 over time, via over-the-air software updates, to deliver incremental value to clients that invest in the platform. In addition, KeepTruckin’s model training becomes increasingly more precise due to its in-house safety team assessing quality in real-time, adding additional risk context and providing input that makes model training and development cycles shorter.

Udit Budhia, director of marketing at Ambarella, said, “For the aftermarket, this is probably the most sophisticated AI vision product for automotive. We’ve seen rapid growth for ADAS warning systems, such as lane departure warnings.” Talking about the benefit of the CV22 SoC in this application, he added, “A key component is the image signal processor (ISP). We support various sensor types including RGGB, RGB-IR, monochrome, RCCB and RCCC, and support HDR processing. We can detect objects in the scene and also have low light filtering, which results in less noise and hence more efficient processing of video. We use less bandwidth and significantly less DRAM.”

On the tools, Budhia said, “Our tool allows development of both front-facing and driver monitoring cameras at the same time.” The programmability and efficiency of Ambarella’s CVflow AI architecture enables significant computer vision performance with very low power consumption, which increases reliability and efficiency to run more models in parallel leading to greater detection and visibility all around. Its complete set of CV tools helps customers like KeepTruckin port their own neural networks and includes support for industry-standard machine learning frameworks such as PyTorch, ONNX, Caffe and TensorFlow.

>> This article was originally published on our sister site, EE Times.


Related Contents:

For more Embedded, subscribe to Embedded’s weekly email newsletter.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.