Back to Blog

FPGA+Nvidia Orin NX+AI Heterogeneous Video and Image Processing Development Platform: Applications in High-End Medical and Industrial Inspection, Supporting Customization and Localization

#fpga开发#图像处理#人工智能

An FPGA+GPU heterogeneous architecture video and image processing development platform, it combines the powerful capabilities of AMD Zynq UltraScale+ MPSoC (FPGA) and NVIDIA Jetson Orin NX (GPU), and is applicable to industries with stringent requirements for image precision and real-time performance.

Zynq UltraScale+ MPSoC XCZU19EG

16 GB

PL-side DDR4

8GB

PS-side DDR4

128MB

QSPI Flash

32GB

eMMC Flash

1143K

Logic Cells

1045K

Flip-Flops

NVIDIA Jetson ORIN

70T/100T

AI Performance

765/918MHz

Max GPU Frequency

6/8 Cores

Arm Cortex-A78AE

2GHz

Max CPU Frequency

32x 1024 Cores

Tensor Cores, NVIDIA Ampere Architecture

Rich Interface Resources

2x

100G Optical Fiber

8x

12G-SDI

……

4x

HDMI 2.1 Input/Output

FMC HPC Interface

Medical Endoscopy and Surgical Robotics Solutions

In applications for medical endoscopy and surgical robotics solutions, it demonstrates its unique advantages. It supports the acquisition, processing, and 8K high-definition video image display, and also provides efficient 3D image processing capabilities. Its advanced synchronization technology ensures left-eye and right-eye video data synchronization accuracy down to 1 line, with end-to-end latency controlled within 50 milliseconds, significantly enhancing real-time performance and response speed. These features provide doctors with clearer, more stereoscopic, and more precise visual information, thereby improving the accuracy and safety of minimally invasive surgery.

Industrial Inspection Solutions

In the field of industrial automation, it also demonstrates its powerful performance. It can achieve efficient object recognition and performance computation in areas such as semiconductor inspection, electronics manufacturing, photovoltaic energy, lithium batteries, and assembly, improving the efficiency and accuracy of production quality inspection. Particularly when using the YOLOX object recognition algorithm, it achieves an mAP of up to 0.9 and a recognition speed of 50 FPS, greatly enhancing visual processing capabilities in industrial automation processes.

Its multi-channel video processing capability, thanks to its 8-channel 12G-SDI bidirectional interfaces and 4-channel HDMI 2.1 input/output, enables its wide application in various fields such as medical imaging, industrial automation, and smart cities & security. Whether for research institutions or high-tech enterprises, it is an ideal choice for product prototyping and testing, providing strong support for imaging requirements across various industries.