GLA Summit 2020/Online Video Compression for ADAS Recording System

From LabVIEW Wiki
Jump to: navigation, search

Online Video Compression for ADAS Recording System by Pravin Ghodake

This is inspired by the non-LabVIEW (VHDL) FPGA library that can be integrated to LabVIEW FPGA for the online camera data compression/decompression. It is an application that we developed that brings us a cost saving of about €600K.

Valeo is a world leader in Advanced Driver Assistance Systems. We provide cameras, ECUs to our customers for the automated parking, lane detection system, etc. We and any company working in this domain put a lot of effort into effectively training and tuning these systems to various road conditions by recording the road video data across the world for different road conditions. And then using this data in HIL/SIL to tune and develop the robust algorithms for these critical automated systems. We typically acquire several petabytes of data for each project which costs millions for its storage on the cloud. For E.g. we want to record 7 petabytes of data over the next 5 years for the current project. Each petabyte costs €233K per petabyte = €1.6M. This data is recorded via NI FlexRIO FPGAs installed in the recording vehicle.

We evaluated and effectively integrated the best video compression FPGA IP (Intellectual Property) from the third party to NI FPGA. This IP must ideally have the smallest footprint, highest compilation frequency, and the highest compression ratio. With this solution being part of our FPGA recording system, we reduce the size of recorded data by about 40%, i.e. to 4.2 petabytes instead of 7 petabytes, i.e. about €600K saving in the cost of data storage. All those working in a similar video recording system using NI FPGA can utilize this solution.

Outline of my presentation:

  1. . Overview of Valeo video recording system
  2. . Overview of the third party video compression IP
  3. . Overview of importing this IP to NI LabVIEW FPGA
  4. . LabVIEW FPGA reusable architecture of video recording with this compression IP

Presentation Links

See Also

External Links