Machine Vision systems

The MRTech company helps its customers develop machine vision solutions to enable customers of any experience and business size to realise their ideas. Based on a deep understanding of how machine vision software works with modern cameras and on computer hardware, including at a low level, we let you maximise the performance of the target system and meet other, sometimes really challenging requirements.

For that MRTech team provides the following services:

  • Project study, recommendation on hardware selection and system architecture
  • Software consulting, assistance with the integration of the IFF SDK
  • Development of custom image processing applications based on the IFF SDK
  • Customization of processing modules and algorithm implementation
  • IT consulting and support
  • Development of custom hardware
  • Study of project feasibility and prototype development

Useful information about this and some examples are provided below.

Your system prototype

 

MRTech has designed a series of prototypes to present the solutions based on the MRTech image processing SDK running on the NVIDIA Jetson platform.

The purpose of each prototype is to provide users with an opportunity to validate the proposed solutions for their tasks, demonstrate the potential of the IFF SDK and evaluate its performance for a specific system configuration. At the same time, it allows us to get useful feedback so that we can help solve the real-world challenges user faces.

Examples of Jetson Xavier NX and AGX Xavier-based prototypes are shown here. Similar systems based on Jetson Orin and other platforms can also be provided.

Xavier NX with USB3 camera

Prototype hardware

  • 3.1 MP XIMEA MC031CG-SY machine vision camera
  • NVIDIA Jetson Xavier NX 16GB module mounted on Auvidea JNX30D carrier
  • Camera and Jetson accessories including cables, lens, power supply, etc.

Installed software

  • L4T operating system and dependencies
  • MRTech IFF SDK, demo version

Provided Xavier NX module is fully programmed to stream images from XIMEA camera via Ethernet using farsight sample from MRTech IFF SDK.

An additional application based on imagebroker SDK sample is installed on the prototype that allows rendering images on a display connected to Jetson. This application can be used to adjust the camera lens and demonstrates how to export images to the user code across IFF SDK library boundaries.

The preset configuration of farsight sample application

  • full-frame acquisition from XIMEA camera
  • color pre-processing on GPU:
    – black level subtraction, histogram calculation, white balance
    – DFPD demosaicing 11×11
    – color correction, enhanced HDR gamma (HLG)
    – image format conversion to YV12 BT709
  • automatic control of exposure time and white balance
  • 264/H.265 encoding, RTSP streaming

Prototype performance when streaming images

  • 3.1 MP 2064×1544 (full frame) @ 12-bit @ 80 FPS
  • Image processing latency on Jetson: 21 ms
  • Expected Glass-to-Glass (G2G) latency: 70-80 ms
  • Jetson GPU engine load: ~ 50%
  • Prototype power consumption: up to 17 W (including camera and Xavier)

Xavier NX with PCIe XIMEA camera

The prototype includes:

  • XIMEA MX089CG-SY-X2G2-FF camera (8.9 MP, PCIe interface)
  • NVIDIA Xavier NX developer kit
  • Camera and Jetson accessories including cables, lens, power supply, etc.
  • Software installed on Jetson:
    – L4T operating system and dependencies
    – Demo version of MRTech IFF SDK

Xavier NX module is fully programmed with a custom application like farsight sample with two preset configurations for Full HD and 4K image processing. See below for the results of the prototype testing.

Full HD (1920 x 1080) 8-bit image at 60 FPS

  • 8-bit image acquisition with 2×2 skipping and 1920×1080 ROI
  • image processing latency on Jetson: 12.4 ms
  • expected G2G latency: ~ 50-60 ms
  • Jetson GPU load: ~ 10%

4K (3840 x 2160) 12-bit image at 50 FPS

  • 12-bit image acquisition with 3840 x 2160 ROI
  • image processing latency: up to 50 ms
  • expected G2G latency: ~ 100-120 ms
  • Jetson GPU load: up to 70%

AGX Xavier with two USB3 cameras

The prototype includes:

  • Two XIMEA MC031CG-SY-UB USB3 cameras with 3.1 MP (2064 x 1544) each
  • NVIDIA Jetson AGX Xavier Developer Kit 32GB
  • Camera and Jetson accessories including cables, lens, power supply, etc.
  • Software installed on Jetson:
    – L4T operating system and dependencies
    – Demo version of MRTech IFF SDK

The prototype user launched a custom application based on imagebroker sample from the IFF SDK. The application receives frames from two cameras and shows the images on the display connected to Jetson.

Image processing on Jetson

IFF SDK benchmarks for Orin

Machine Vision systems

Software for Machine Vision systems

If you haven’t decided which software to use in your machine vision system or would like to explore alternatives to your current solution, we suggest you look at the cross-platform image processing SDK from MRTech SK.

The MRTech IFF SDK is designed to deliver images to the user’s application in the most efficient way possible. This means that you (your application or library) get the images with minimal latency, in the right format and quality. The SDK runs on all major platforms – Linux, Windows, Jetson, and supports input from many machine vision cameras with PCIe, USB3, Ethernet, MIPI interfaces.

Up-to-date documentation for the IFF SDK as well as simple applications for using it are now available on our company’s Github account.

If necessary, we will provide you with the last Trial version of the IFF SDK on request.

IFF SDK for image processing

MRTech on Github

IFF SDK online demo