<- back to shop

SGM®Producer

The SGM®Producer is a software library for running stereo matching processes on the graphics card of an external computer – particularly useful when high-speed processing is of essence.

  • For applications that need high-quality depth images at high frame rates
  • Frame rate of depth images up to 25 Hz
  • GenICam-compatible producer
  • No change of interface as compared to on-board stereo matching
  • Processing for multiple sensors on one GPU is possible
Categories: ,

Description

The stereo matching component is part of the basic on-board software package on each rc_visard. The SGM®Producer is an add-on software element that enables a significant speed-up of the depth image computation by utilizing the computing power of an external host computer’s graphics card (GPU) for stereo matching. Frame rates of 25Hz can be achieved.

Switching from on-board to off-board stereo matching is easy: All interfaces, application code, Semi-Global Matching (SGM) processing and the achieved results are identical – the key difference is the processing speed.

As an example, on an Nvidia GeForce 1080, depth images can be computed with 25 Hz in low, medium and high quality, even from two rc_visards at the same time. Full-quality stereo matching in 1.2 MPixel resolution is possible with 8 Hz on this graphics card (as compared to 0.8 Hz on-board the sensor).

The software library implements the GenICam transport layer interface standard and can be used with all GenICam-compatible applications that support the current GenTL 1.5 standard, e.g. Halcon from MVTec (just replace the GigE Vision producer by Roboception’s library!).


Downloads and Links

SGM®Producer Software Pack (dongle required for activation)
SGM®Producer Manual (.pdf)
rc_visard GenICam API

HALCON users: Note that a sample program is available within the latest HALCON release!

Additional information

Software Interface

rc_visard GenICam API or other compatible application (GenICam GenTL version 1.5 compatible)

Required Host OS

Windows 10 (64 bit), Ubuntu 18.04 (64 bit) or Ubuntu 16.04 (64 bit). Other operating systems on request.

Required Graphics Card

Nvidia GPU with computing capability of 3.5 (i.e. Kepler architecture) or higher, with a minimum of 1.5 GB memory per rc_visard.

Required Interfaces

Gigabit Ethernet network port for rc_visard and free USB port for license dongle

You may also like…