Leopard Imaging will use VISION 2026 in Stuttgart to demonstrate a broad set of camera and AI perception systems aimed at robotics, industrial automation, autonomous machines and edge AI. The portfolio spans long-range 3D sensing, stereo RGB-IR cameras, multispectral imaging, thermal perception and high-resolution industrial vision, with several systems built around NVIDIA platforms and high-speed connectivity.
Computer vision is moving increasingly close to the point where machines interact with the physical world. For robots, autonomous systems and industrial equipment, that shift requires more than high-resolution cameras. Vision systems must combine sensing, synchronization, depth information and increasingly local AI processing while operating within tight latency and power constraints.
Leopard Imaging is bringing that combination to VISION 2026, scheduled for October 6–8 in Stuttgart, Germany. At Hall 10, Booth 10G79, the company plans to demonstrate camera and AI perception technologies covering robotics, industrial inspection, machine vision and autonomous systems.
The portfolio includes 3D perception, RGB-IR stereo imaging, multispectral cameras, thermal sensing and high-resolution industrial imaging. Several demonstrations also highlight the growing role of edge AI platforms in processing sensor data closer to where it is captured.
3D perception moves toward software-defined sensing
One of the showcased systems is Sirius Max, a compact RGB-plus-depth platform combining intelligent indirect Time-of-Flight (iToF) sensing, programmable beam-steered illumination and embedded synchronization.
The system uses Lumotive’s LCM technology for beam steering and is designed for long-range 3D perception. Its software-defined scanning approach is aimed at robotics and autonomous systems that need to interpret spatial information dynamically.
This type of sensing is becoming increasingly important as robots move beyond fixed industrial environments. Autonomous machines need to detect objects, understand their surroundings and build spatial representations in real time, making depth cameras and 3D sensors important components of AI perception stacks.
Stereo cameras combine RGB, infrared and edge computing
Leopard Imaging’s ASTRO Stereo Camera takes another approach to 3D perception. The system combines two 5.1-megapixel BSI global-shutter sensors with IR projection and illumination.
The camera is being demonstrated with NVIDIA Holoscan 10GigE connectivity and the NVIDIA Jetson Thor platform. It supports synchronized RGB and IR stereo imaging, depth perception, 3D reconstruction and AI perception pipelines.
The architecture reflects a broader movement toward high-bandwidth sensor-to-compute connections. Instead of treating a camera as an isolated imaging peripheral, industrial vision systems increasingly function as components within distributed AI computing architectures.
NVIDIA’s Holoscan platform is designed to process high-volume sensor data for real-time AI applications, while Jetson platforms target edge computing workloads. That combination is particularly relevant for robotics, where sending every camera frame to a remote server can introduce latency and connectivity constraints.
Industrial inspection demands resolution and processing
Leopard Imaging will also demonstrate an IMX530-based camera using Sony’s 25-megapixel global-shutter sensor.
The NVIDIA Holoscan 10GigE solution combines high-resolution imaging with FPGA processing. Target applications include industrial inspection, barcode recognition and logistics automation.
Global-shutter sensors are particularly useful for machine vision because they capture an image simultaneously across the sensor, reducing motion distortion when objects move through a production or logistics environment.
The combination of high resolution and local processing is also relevant as manufacturers use computer vision for increasingly detailed inspection tasks. AI models can identify defects or classify objects, but the quality and consistency of the image pipeline remains a fundamental part of system performance.
Hybrid depth sensing addresses interference
Another demonstration focuses specifically on depth accuracy.
Leopard Imaging’s Hybrid iToF RGB-D Camera combines pulsed and phased indirect Time-of-Flight technologies with a 1.2-megapixel onsemi depth sensor. The company positions the architecture for applications including robotics, 3D mapping, gesture recognition and augmented and virtual reality.
Using multiple ToF techniques can help address some of the interference challenges associated with depth sensing. That becomes important when several sensors operate within the same environment, as can happen in warehouses, factories or multi-robot deployments.
Multispectral and thermal cameras broaden machine perception
The company’s IMX454 multispectral camera extends machine vision beyond conventional visible-light imaging. It captures 41 spectral channels across wavelengths from 450 to 850 nanometers, with a resolution of 1,948 × 1,097 pixels at 30 frames per second.
The USB 3.0 UVC design targets applications including agriculture, food inspection, environmental monitoring and spectral analysis.
Multispectral imaging can provide information that conventional RGB cameras cannot capture, allowing AI systems to distinguish materials or conditions based on their spectral signatures. In agriculture and food inspection, for example, those additional signals can contribute to classification and quality-control workflows.
Leopard Imaging is also showing DragonFire, an uncooled infrared camera designed for thermal perception. It provides 640 × 512 resolution at 30 or 60 frames per second, with 12-micrometer pixels and stated NETD performance of 40 mK or lower.
Its GMSL2 interface targets robotics, industrial inspection, autonomous systems and thermal monitoring.
Thermal sensing gives autonomous and industrial systems another source of information when visible-light cameras may be insufficient, particularly for detecting temperature differences or operating in challenging lighting conditions.
RGB-IR cameras target active stereo applications
Two additional systems emphasize RGB-IR perception.
Eagle 1 combines dual 5.1-megapixel BSI global-shutter RGB-IR sensors with active IR illumination and a dot-pattern projector. It produces RAW RGB and infrared data for depth perception, navigation, robotics and autonomous machines.
Meanwhile, Dragonfly USB 3.0 YUV combines an onsemi 4K HyperLux LP sensor with an integrated image signal processor, HDR imaging and USB 3.0 connectivity. Its plug-and-play design is aimed at AI vision development, smart kiosks, video conferencing and industrial applications.
Edge AI increasingly depends on the sensor pipeline
The demonstrations at VISION 2026 point to a larger trend in AI perception platforms: computer vision is becoming a coordinated system of sensors, interfaces, processing hardware and AI models.
The market for edge AI is expanding as manufacturers seek to process data locally for latency, privacy and operational reasons. Gartner forecasts worldwide AI spending at approximately $2.7 trillion in 2026, with AI infrastructure accounting for a substantial portion of that investment.
For robotics and industrial automation, however, AI infrastructure starts before the accelerator. The camera determines what information reaches the model, while synchronization, depth accuracy, dynamic range, thermal sensitivity and data bandwidth can determine how reliably the system perceives its environment.
Leopard Imaging’s VISION 2026 portfolio reflects that convergence. Rather than focusing on a single sensor category, the company is demonstrating a collection of imaging technologies designed for different AI perception workloads.
The practical challenge for manufacturers will be integrating those sensing systems into reliable production pipelines. As robots, autonomous machines and smart factories become more dependent on real-time perception, the camera is increasingly becoming an active component of the AI infrastructure rather than simply a source of images.
Market Landscape
AI-powered machine vision is expanding across robotics, manufacturing, logistics, agriculture and autonomous systems. The shift toward edge processing is increasing demand for high-bandwidth interfaces, global-shutter sensors, depth cameras, multispectral imaging and thermal perception.
NVIDIA’s edge computing ecosystem, including Jetson and Holoscan, is one part of a broader industry movement toward processing sensor data locally. Camera vendors are consequently competing not only on resolution, but also on synchronization, embedded processing, AI compatibility and the ability to integrate into real-time perception pipelines.
Top Insights
- Leopard Imaging will demonstrate 3D, stereo, multispectral and thermal cameras at VISION 2026 in Stuttgart.
- Sirius Max combines RGB, iToF depth sensing and programmable beam steering for long-range 3D perception.
- ASTRO pairs RGB-IR stereo imaging with NVIDIA Holoscan 10GigE and Jetson Thor for AI perception workloads.
- IMX530 and DragonFire target demanding industrial vision, logistics, autonomous-system and thermal applications.
- The portfolio illustrates how sensors, edge computing and AI models are converging into integrated perception systems.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
