Topic

Automotive

The FPGA design of driving assistance features, such as Lane Departure Warning, Pedestrian Detection and Vehicle Detection and their integration into an automotive driving assistance development platform based on AMD Zynq-7000 SoC and Zynq UltraScale+ MPSoC devices.

«I’ve worked with EVS Embedded Vision Systems for nearly a decade. They are one of my trusted ecosystem partners for automotive-related solutions development. I have relied on the EVS team to do research and produce proof-of-concept feasibility studies – particularly in the area of image processing. Furthermore, EVS has proven itself very capable in creating designs, including FPGA IP cores, that can be used for demonstration and commercialized for production use. In conducting projects with the EVS team they have always maintained a high level of professionalism and are exemplary in communicating project progress and technical detail»

― Paul Zoratti

Chief System Architect – Automotive Adaptive Compute | AMD

Problem

Cameras are a critical element in AD and ADAS

Cameras are a critical element in Automated Driving (AD) and Advanced Driving Assistance Systems (ADAS) because they provide the advanced sensing capabilities required for vehicle perception, control, and path planning.

Automotive system integrators and OEM car makers are obliged to prototype new technologies, carry out on-board testing, customize and quickly integrate them with different sensors and do feasibility studies and proofs of concept. This must be done in a very short time so as not to lose business opportunities and to reduce the time to market.

Solution

Multi-camera embedded vision platform

EVS has collaborated at length with AMD Inc., Xylon d.o.o. and other AMD partners to build up a multi-camera embedded vision platform for the quick development of ADAS solutions based on the AMD Zynq SoC.

This platform, called logiADAK, is a complete out-of-the-box development platform that includes application-specific software and a customizable reference design which can be used for rapid test vehicle installation and demonstration. Over the years, it has evolved to adapt to new AMD devices, new development boards and additional support functions.

Added value

A multitude of ready-to-use features

LogiADAK means that numerous automotive companies worldwide can experiment with DA functions on AMD SoC leveraging a platform offering a multitude of ready-to-use features that are easy to customize and integrate on board the vehicle.

EVS contribution to the Automotive Driving Assistance Development Kit

In the ADAK framework, EVS has developed and integrated several vision-based driving assistance functionalities that are hardware-accelerated due to proprietary IP cores that use image processing, computer vision and machine learning tasks. See the table below.


LogiHOG IP CORE

Fundamental building block for pedestrian and motorbike detection.

  •  Advanced HOG/SVM based algorithm
  • Trained on a wide range of roadway scenarios using STK software

LogiVDET IP CORE

Fundamental building block for vehicle detection to support Adaptive Cruise Control and Forward Collision Warning.

  • Classifier based on Local Binary Pattern and cascaded classifiers
  • Car and truck detection
  • Range Estimation (5 to 100m)

LogiLMD IP CORE

Fundamental building block for a rear-looking Lane Departure Warning system.

  • The core works in the top view space
  • Lane marking detection based on image processing and fitting of geometric primitives
Project partner

AMD is a world leader in programmable logic devices. EVS has been member of the AMD Adaptive Computing Partner Program since 2009.

Xylon is a leading provider of proven FPGA application solutions and cutting-edge IP cores optimized for AMD® FPGAs in embedded graphics, video, image processing and networking.

Insights

Fully Integrated ADAS

Fully Integrated ADAS: Cost-effective, scalable, multi-camera, multi-function, driver assistance platform

  • Multiple integrated DA features in a single platform:
    • 3D 360o Surround View
    • Pedestrian Detection
    • Vehicle Detection
    • Lane Departure Warning
    • Blind Spot Detection
    • Rear View Camera
  • Flexible on-chip partitioning for multiple parallel video processing pipelines to support several ADAS features simultaneously
  • Licensable IP Cores and expansion room for your ADAS algorithms