As real-world applications across outdoor, industrial, and automotive sectors place higher demands on thermal imaging, the technology is evolving from capturing and presenting infrared information toward analyzing and understanding thermal data. At CIOE 2026 in September, Raytron officially launched Falcon500, its third-generation infrared AI image processing chip, integrating infrared AI-ISP, intelligent sensing, multimodal fusion, and precise temperature measurement into thermal imaging systems, further advancing thermal imaging from seeing to understanding.
An infrared AI image processing chip (infrared AI-ISP ASIC) is a new generation of single-chip intelligent imaging solution specifically designed for thermal imaging applications.
Due to the physical characteristics of infrared detectors, raw infrared image signals inherently exhibit issues such as temperature drift and non-uniformity. Dedicated ISP (Image Signal Processor) architectures are therefore required to perform non-uniformity correction, noise reduction, and image restoration. These image processing capabilities are typically implemented by dedicated image processing chips, which further convert raw signals from the detector into clear and stable thermal images.
Building on this foundation, traditional infrared image processing chips have primarily focused on image generation. As AI enters the thermal imaging industry, processing capabilities are expanding from “outputting images” to “understanding images.” AI can not only further enhance image quality and scene adaptability, but also enable thermal imaging systems to support target recognition and intelligent sensing.
Compared with general-purpose chips such as CPUs and GPUs, ASICs (Application-Specific Integrated Circuits) are architecturally optimized for specific applications, enabling higher processing performance, greater integration, and improved energy efficiency. With high-volume production, they can also offer significant cost advantages.
Infrared AI-ISP is not simply a replacement for conventional image processing. It is specifically optimized for the characteristics of infrared thermal imaging.
Raytron’s in-house Falcon500, its third-generation infrared AI image processing chip, goes beyond the capabilities of traditional infrared ISP by integrating high-performance computing, a dedicated infrared AI-ISP architecture, intelligent recognition, multimodal fusion, and precise temperature measurement into a single chip. It also enhances product adaptability, with five core capabilities:
Falcon500 combines a high-performance CPU with floating-point computing and large-capacity cache to improve the efficiency of image processing and temperature calculations. High-speed storage and memory provide sufficient computing resources for AI algorithms. Built-in hardware video compression acceleration reduces data size at the same image quality, lowering bandwidth requirements for remote transmission. Falcon500 supports ultra-high-resolution image output and maintains smooth performance even when multiple video streams are transmitted simultaneously, maximizing the combined capabilities of high-quality imaging and intelligent analysis.
Built on a new-generation imaging processing architecture, Falcon500 performs integrated optimization of image details, noise, and textures to suppress artifacts such as water ripples, smearing, and abnormal noise while preserving natural textures. It also applies multi-mode correction strategies optimized for complex conditions, including motion, scene changes, and large temperature differences, reducing motion trails and image flicker for more stable imaging in dynamic scenes.
Falcon500 is trained on tens of millions of real-world samples to fully leverage the capabilities of its hardware and improve recognition accuracy. Its in-house heterogeneous NPU is optimized through hardware-software co-design to efficiently support proprietary detection models and meet real-time inference requirements.
Pixel-level image alignment enables deep fusion of visible-light details and infrared temperature information. Dynamic power management adapts to different operating conditions while balancing performance and battery life. Falcon500 can also integrate up to five types of video streams, distortion correction, high-speed interfaces, hardware encryption, and fault protection, with additional support for color CVBS, MIPI, DSI, and IN interfaces.
Falcon500 restructures the entire temperature measurement pipeline to address inaccurate measurements and temperature fluctuations. Its optimized algorithms and flexible KT quantization mechanism balance quantization precision and measurement range, while algorithmic correction for optical hole effects reduces measurement deviations caused by differences in target size. Environmental correction and isothermal noise reduction are also implemented in hardware to reduce measurement latency and improve data stability, while simplifying mass-production calibration and shortening system development cycles.
By integrating multiple core capabilities, Falcon500 further enhances the processing and scene adaptability of thermal imaging systems, supporting the needs of different end products and system solutions for image processing, intelligent analysis, and precise temperature measurement. Modules powered by Falcon500 can deliver clearer and more stable images while using AI processing to extract more information about targets and scenes from infrared data.
The following highlights showcase Falcon500’s performance in thermal imaging applications across industrial inspection, outdoor observation, automotive driver assistance, and low-altitude applications:
Industrial Temperature Measurement: Pixel-level image alignment enables deep fusion of visible-light details and infrared temperature information. Combined with a dedicated temperature measurement processing pipeline and environmental correction capabilities, Falcon500 provides stable temperature measurement for precision industrial inspection and equipment condition monitoring.
Outdoor Applications: Multiple packaging options support compact designs with higher integration and lower power consumption, while combining high-resolution imaging, precise temperature measurement, and AI capabilities to meet the requirements of portable devices for lightweight design and overall performance.
Driver Assistance: Advanced noise reduction and high-dynamic-range imaging help suppress road-scene noise and motion artifacts such as trailing and flickering. AI-based target recognition enables both imaging and intelligent analysis, providing an on-chip processing foundation for pedestrian detection, vehicle recognition, and intelligent warning functions.
Low-Altitude Applications: High-resolution imaging and AI super-resolution enhance the visibility of distant small targets. The low-power architecture balances high-quality detection with battery life, making it suitable for long-endurance platforms such as commercial and industrial drones.
As a leader in infrared thermal imaging, the launch of Falcon500 represents another advancement in Raytron’s AI-powered sensing technologies. By integrating high-quality imaging, precise temperature measurement, and on-device AI recognition into a single chip, Falcon500 helps lower the R&D and manufacturing barriers for thermal imaging products, enabling equipment manufacturers to develop fully featured thermal imaging systems in shorter cycles and with more predictable costs. This helps accelerate the adoption of intelligent thermal imaging across applications including power inspection, security and fire protection, automotive, industrial inspection, and low-altitude operations.
