As infrared thermal imaging expands into outdoor, industrial, automotive, and emerging commercial low-altitude applications, the demand for higher image quality and more advanced image analysis continues to grow. On September 9, 2026, at the China International Optoelectronic Exposition (CIOE), Raytron OEM unveiled Falcon500, its third-generation infrared AI image processor. The new chip brings major improvements in image quality, intelligent sensing, multimodal integration, and precision temperature measurement, further expanding the processing capabilities of thermal imaging systems.
Why Dedicated AI Processing Matters for Thermal Imaging
A dedicated infrared AI image processing chip, or infrared AI-ISP ASIC, is an application-specific processing chip designed for thermal imaging applications.
Due to the inherent physical characteristics of infrared detectors, raw signals can be affected by temperature drift, non-uniformity, and other issues. A dedicated ISP, or Image Signal Processor, is therefore required to perform functions such as non-uniformity correction, noise reduction, and image restoration to produce clear, stable thermal images.
The addition of AI further extends the capabilities of the system. In addition to improving image quality and scene adaptability, AI enables more advanced intelligent sensing functions. Compared with general-purpose processors such as CPUs and GPUs, an ASIC—an application-specific integrated circuit designed for a particular task—can offer higher processing performance, a higher level of integration, and better energy efficiency. It can also reduce system costs in high-volume production.
Put simply, a dedicated infrared AI image processing chip converts raw data from an infrared detector into thermal images that can be viewed, analyzed, and used for temperature measurement. It combines image processing and AI capabilities to turn raw detector signals into usable image and temperature data.
Dedicated Infrared AI Image Processing Chip — Falcon500
Falcon500 is Raytron OEM’s self-developed third-generation dedicated infrared AI image processing chip.
Building on conventional infrared ISP architectures, Falcon500 integrates five core capabilities: High-Performance Computing Platform, Professional Infrared AI-ISP, Intelligent Sensing, Multimodal Integration, and Precision Temperature Measurement.
High-performance CPU Processing
Falcon500 combines strong CPU performance with floating-point processing and large-capacity cache, improving the efficiency of image processing and temperature calculations. High-speed memory and storage provide the resources required for AI processing and real-time inference.
Hardware-accelerated video compression reduces bandwidth requirements while preserving image quality, supporting efficient remote transmission. Falcon500 also supports ultra-high-resolution image output and smooth multi-stream transmission, allowing high-resolution imaging and intelligent analysis to run efficiently in parallel.
Professional Infrared AI-ISP
Built on a new-generation image processing architecture, Falcon500 provides integrated optimization of image detail, noise reduction, and texture preservation. It helps reduce ripple artifacts, smearing, and abnormal noise while preserving natural textures, improving edge clarity, and delivering smoother transitions between bright and dark areas.
A multi-mode correction strategy is used to optimize imaging for complex conditions such as motion, scene changes, and large temperature differences. This helps reduce motion artifacts and image flicker while improving imaging stability in dynamic scenes.
Intelligent Sensing
Falcon500’s AI capabilities are developed using a dataset containing tens of millions of real-world samples, helping improve NPU-based recognition accuracy.
At the same time, Raytron OEM’s self-developed heterogeneous NPU uses hardware-software co-optimization to efficiently run in-house detection models for real-time inference.
Multimodal Integration
Falcon500 uses pixel-level image alignment to combine visible-light details and textures with infrared temperature information.
Dynamic power management allows the chip to adapt to different operating conditions while balancing performance and power consumption. Falcon500 supports the integration of up to five video streams of different types, along with system functions such as distortion correction, hardware encryption, fault protection, and high-speed interfaces. The addition of color CVBS, MIPI, DSI, and IN interfaces further expands integration flexibility.
Precision Temperature Measurement
Falcon500 improves the entire temperature measurement processing chain to address common issues such as measurement errors and fluctuating temperature readings.
Its optimized temperature measurement algorithms use a more flexible KT quantization mechanism to achieve a better balance between quantization accuracy and measurement range. Algorithmic correction for optical aperture effects helps reduce measurement errors associated with variations in target size.
Functions including environmental correction and isothermal noise reduction are implemented directly in hardware, further reducing measurement latency and improving data stability. These improvements also simplify calibration during volume production, helping accelerate product development and shorten time to market.
Falcon500 Enhances Product Capabilities Across Multiple Applications
Outdoor Applications
Falcon500 offers multiple packaging options to support compact product designs, delivering a higher level of integration and lower power consumption while combining high-resolution imaging, precision temperature measurement, and AI capabilities.
Compared to the previous generation, the Falcon500 delivers notable improvements in image quality, with a Mean Opinion Score (MOS) increase of 15% in typical scenes and 30% in challenging scenes, making it well-suited for compact, portable thermal imaging and inspection devices.
Low-Altitude Applications
Falcon500 supports ultra-high-resolution imaging and AI super-resolution. Its DCL algorithm improves image clarity for small objects viewed over longer distances.
The chip’s low-power architecture also balances high-resolution imaging performance with operating endurance, making it suitable for civilian low-altitude platforms that require longer operating times.
Industrial Temperature Measurement Applications
Pixel-level image alignment enables visible-light details and textures to be fused with infrared temperature information.
A dedicated hardware-based temperature measurement pipeline works together with optical correction and intelligent environmental compensation to reduce the effects of environmental variations and temperature drift, helping maintain measurement accuracy and stability. These capabilities are suitable for applications such as precision industrial inspection and equipment condition checks.
Driver Assistance Applications
Advanced noise reduction combined with high-dynamic-range imaging helps reduce road-scene noise, motion artifacts, and image flicker.
AI-based object recognition works alongside image processing to support functions such as pedestrian detection, vehicle recognition, and intelligent alerts, providing an integrated on-chip processing foundation for driver-assistance applications.
Conclusion
Falcon500 integrates three core capabilities—high-resolution imaging, precision temperature measurement, and on-device AI recognition—into a single chip. This integration can simplify thermal imaging product development, helping manufacturers shorten development cycles, manage costs, and deliver more capable products.
With continued advances in its underlying chip technology, Raytron OEM aims to enhance product capabilities across applications including power equipment inspection, fire prevention, in-vehicle systems, industrial inspection, and civilian low-altitude applications.