1 Overview

High-performance embedded High-Speed Image Acquisition and Edge AI platform for industrial and scientific research scenarios

In industrial inspection, scientific research or high-speed imaging scenarios, users often face the following challenges:

1. What is Camera Link?

→ Camera Link is a widely adopted high-speed digital image interface standard in machine vision, maintained by the AIA (Automated Imaging Association). Designed for industrial cameras, scientific cameras, line scan cameras and high-speed acquisition systems, it features the following strengths:

1) Stable transmission links with mature industrial deployment;

2) Fixed low latency, ideal for hard real-time systems;

3) Professional frame grabber receivers support continuous high-bandwidth data reception;

4) Remains a critical connection standard for cameras, frame grabbers and processing platforms in numerous industrial projects.

2. Massive image data generated by Camera Link cameras cannot be analyzed in real time (e.g. defect detection, target tracking) due to limited CPU processing speed?

→ GPU is the optimal solution. The NVIDIA Jetson GPU acts as a "visual brain", processing thousands of pixels in parallel. It runs AI Models (defect detection, classification, etc.) dozens of times faster than traditional processors with ultra-low power consumption, perfect for embedded deployment.

3. All-in-one demand: high-speed acquisition, intelligent analysis, compact form factor and multi-user remote real-time monitoring?

→ Conventional solutions require multiple discrete hardware: frame grabber + industrial PC + encoder + network module...

The JetKU-CL4001 integrates all functions onto a single board:

FPGA: High-speed buffer for raw data from Camera Link cameras;

GPU: Real-time visual content analysis & AI inference;

Live video broadcast to remote clients via standard RTSP protocol with one-click configuration.

In short: JetKU-CL4001 = High-speed camera receiver + Visual AI processing core + Video streaming transmitter.

Typical Application Scenarios: UAV low-altitude defense, Semiconductor Wafer Defect Inspection, edge computing for high-end medical imaging, intelligent sorting & quality inspection on high-speed production lines

JetKU-CL4001 is an embedded image acquisition board tailored for high-frame-rate, high-data-volume industrial vision and scientific imaging applications. It adopts a heterogeneous computing architecture combining Xilinx Kintex UltraScale+ FPGA and NVIDIA Jetson GPU, interconnected via PCIe Gen3 x4. While supporting high-speed image capture from Camera Link cameras, it delivers powerful real-time preprocessing, AI Inference and low-latency RTSP video streaming capabilities.

Paired with StarTest Electronics self-developed ImgGrab graphical host software, users can rapidly configure cameras, capture and preview images. Supported by the integrated software stack of Aravis + GStreamer + MediaMTX, the JetKU-CL4001 streams captured frames over standard RTSP to remote terminals or video management platforms, forming a complete all-in-one edge vision solution covering "acquisition — processing — distribution".

2 RTSP Real-Time Streaming Capability

The JetKU-CL4001 works not only as an image capture device, but also as an intelligent edge video node. Built upon open-source ecosystem components, it seamlessly pushes Camera Link camera feeds over standard RTSP to local LAN or cloud servers:

2.1 Integrated Software Stack

(1) Aravis: GenICam-compliant camera control library to retrieve raw image frames from the underlying FPGA driver layer;

(2) GStreamer: Construct low-latency high-efficiency video processing pipelines (format conversion, scaling, encoding);

(3) MediaMTX: Lightweight RTSP/RTMP server supporting multi-stream concurrent publishing with H.264/H.265 encoding.

3 End-to-End Workflow: Image Acquisition to RTSP Streaming

(1) Connect Camera Link Camera

Link the camera to the FMC daughter card via standard cables and supply external power to the camera.

(2) Boot the Jetson System

The system automatically loads the FPGA driver and Aravis GenTL Producer.

(3) Verify Capture via ImgGrab

Launch ImgGrab to confirm camera detection and normal image preview.

(4) Start the MediaMTX Service

Regarded as a high-performance streaming switcher, it interconnects various streaming devices and protocols for cross-protocol media transmission.

(5) Execute GStreamer Streaming Command

Adopt the provided command template, specify camera name and resolution;

The pipeline automatically completes full workflow: image capture → encoding → RTSP streaming.

(6) Remote Access to RTSP Stream

Open network stream in VLC with URL: rtsp://192.168.0.169:8554/live/camera1