3D Around View Monitoring + Parking Detection (3D AVM)

A proprietary 3D Around View Monitoring (3D AVM) system that takes the front, rear, left, and right fisheye-camera images, de-warps and aligns them, then projects and fuses them into a unified 3D model of the vehicle’s surroundings, for perception and automatic parking.

Core techniques include fisheye-camera calibration, image projection and seam stitching, and real-time 3D reconstruction and rendering on embedded platforms (Fujitsu, Nvidia) using OpenCV and OpenGL ES. The system was designed to run on any embedded platform and extended to parking-space detection integrated into automatic-parking products.

The technology shipped through two delivery tracks:

• A Fujitsu-platform version delivered to a Thai car manufacturer, entering its supply chain as a Tier 1, automotive mass-production-grade delivery.
• A self-developed version with Parking-Lot Detection (PLD), flagging available parking spaces on the bird’s-eye view in real time to support automatic parking; this version was handed off via technology transfer to Whetron China (the Suzhou subsidiary), who integrated and delivered it to their end customer.

Together they demonstrate mass-production-grade software delivery that meets strict automotive requirements and transfers cleanly across platforms and teams. (A technical write-up of the self-developed AVM+PLD is linked above; source code is private.)