High Performance meets Ultra-Low Power In Vision ADAS on S32V MCUs

Document created by edk Employee on Apr 5, 2016Last modified by Diana Torres on Apr 5, 2017
Version 5Show Document
  • View in full screen mode


High performance feature extraction and tracking application at ultra-low power on S32V platform.

This demo showcases a real-time high computation algorithm with image capture and display running on a portion of the resources available on the S32V234. Customers can create demanding Automotive grade vision systems such as stereo and single camera as well as advanced surround view systems based on this demo.

The application was written using APEX-CV pro library and demonstrates that high performance application leveraging the APEX Image Cognition Processor cores of the S32V234 could also be easy to write




The APEX cores, with a combined 128 parallel computational units, crunch numbers quickly and at a fraction of the power. Fully programmable, the cores can execute standard and/or customized vision algorithms for ADAS applications and beyond.

The S32V234 MCU captures raw images from HD sensor, and then formats the images with its on-chip ISP that here provides exposure control, white balancing, RGB to Y color conversion.  Formatted images are then feed into the APEX cores that generate multi-level image pyramids, and combined Harris Corner for feature detection followed by Lukas-Kanade (KLT) Sparse Optical Flow for feature tracking.  Then features and displacement are overlaid on image and displayed, at the processing performance of up to 100 fps



NXP Recommends

The S32V230 Processor family for Vision ADAS, includes the award winning automotive grade S32V234 MCU with dual APEX Image Cognition Processor cores.



Video Links


1 person found this helpful