Two Different Physics
Night vision and thermal imaging are fundamentally different technologies that detect different parts of the electromagnetic spectrum. Night vision amplifies ambient visible and near-infrared light (0.4-0.9 micrometers). Thermal imaging detects emitted heat radiation in the mid-wave (3-5 um) or long-wave (8-14 um) infrared bands. Each has capabilities the other lacks, and modern tactical systems increasingly fuse both into a single display.
Image Intensification (Night Vision)
Image intensifier tubes (I2) amplify available light. Moonlight, starlight, or ambient urban illumination. By 20,000-50,000 times through a photocathode-microchannel plate-phosphor screen cascade.
Generation Progression
- Gen I: S-20 photocathode. 120 lp/mm resolution. Bloom from bright sources. Largely obsolete for military use.
- Gen II: Microchannel plate added. 280-380 lp/mm. Better low-light performance. Still in wide use outside the U.S.
- Gen III: GaAs photocathode with ion barrier film. 64-72 lp/mm system resolution. Sensitivity: 1800-2800 uA/lm. The U.S. Military standard since the 1990s.
- Gen III thin-film (WP): White phosphor output replacing green. Improved contrast perception and reduced eye fatigue. Same sensitivity, better image quality perception.
- Gen III unfilmed: Removed ion barrier film for higher sensitivity (2200+ uA/lm) at the cost of reduced tube life. Used in ENVG systems.
Thermal Imaging
Thermal imagers detect emitted infrared radiation from objects based on their temperature. They require zero ambient light. Complete darkness, fog, smoke, and dust do not significantly affect performance. Two primary detector technologies:
- Uncooled microbolometer: Vanadium oxide (VOx) or amorphous silicon (a-Si) resistive elements that change resistance with temperature. NETD: 30-50 mK. No cryogenic cooling required. Weight: 100-300g for a helmet-mounted sensor. Resolution: 320x240 to 640x512 pixels. Frame rate: 30-60 Hz. Cost: $2,000-15,000.
- Cooled semiconductor: InSb or MCT (HgCdTe) photodetectors cooled to 77K by Stirling or pulse-tube coolers. NETD: 10-20 mK. Superior sensitivity and range. Weight: 500g-2kg. Resolution: 640x512 to 1280x1024. Cost: $50,000-200,000. Currently too heavy and expensive for helmet mounting; used in vehicle and weapon sights.
Tactical Comparison
Sensor Fusion: ENVG-B
The Enhanced Night Vision Goggle - Binocular (ENVG-B, AN/PSQ-20B) fuses Gen III image intensification with an uncooled thermal sensor in a single binocular housing. The operator sees an overlaid image combining the high-resolution detail of I2 with the threat detection capability of thermal.
Fusion modes include:
- Outline fusion: Thermal signatures overlaid as edge highlights on the I2 image. Hot objects (people, vehicles) appear with a visible edge glow in the otherwise natural-looking I2 scene.
- Blended fusion: Pixel-level combination of I2 and thermal data. The image shows both detail (from I2) and thermal contrast simultaneously. More information but can be visually confusing.
- Picture-in-picture: Thermal displayed in a small window within the I2 field of view. Operator gets full I2 scene with a dedicated thermal window for threat scanning.
- Toggle: Rapid switching between I2 and thermal with a single button press. Operator selects the sensor best suited to current conditions.
Emerging Technologies
- SWIR (Short-Wave Infrared): 1.0-1.7 um band. Sees through fog better than I2, provides natural-looking images, and works in complete darkness with covert SWIR illuminators. InGaAs sensor costs are dropping rapidly.
- Digital I2: CMOS sensors replacing traditional tubes. Infinite tube life, digital processing, lower cost. Currently lower sensitivity than Gen III tubes but improving each generation.
- AI-enhanced fusion: Machine learning algorithms that automatically select and blend I2, thermal, and SWIR channels based on scene content. Highlights threats, suppresses clutter, adjusts fusion parameters in real-time without operator input.
The debate over night vision versus thermal is obsolete. The answer is both, fused intelligently. The question is how to present two data streams to a human brain that evolved for one. That is the interface problem sensor fusion must solve.