# Thermal Camera Motion Blur on Drones, Vehicles and Trains Why a thermal core smears on drones, vehicles and trains, why a higher frame rate does not fix it, and at what speed temperature readings drop (calculated). Source: https://thermal.torsuntech.com/resources/thermal-camera-moving-platform-blur/ [Resources](https://thermal.torsuntech.com/resources/) / Moving platforms # Thermal cores on drones, vehicles and trains: when the image smears and when the temperature reads low “Will it blur on a drone, a car or a train, and do we need a higher frame rate?” The short answer: the smear comes from the detector, not the frame rate, and the temperature reading goes wrong before the image looks bad. Here is the arithmetic, with the 640 × 512 radiometric core we sell as the example. Every speed and percentage below is calculated with a simple first-order model, not measured. ## Why frame rate is not the limit An uncooled microbolometer pixel does not take a snapshot. It warms and cools with a thermal time constant, τ. On this core τ is under 12 ms; uncooled cores typically sit around 10–15 ms. The core delivers 30 frames per second, but each frame still carries what the pixels did over the last few time constants. A higher frame rate gives you more frames per pass, not sharper frames. ## How long is the smear? Looking sideways from a moving platform, the smear on the target is roughly speed × τ. With τ = 12 ms: - 15 km/h: about 50 mm - 36 km/h: about 120 mm - 80 km/h: about 270 mm This does not depend on distance or lens, and neither does the ratio of smear to target size. A 50 mm part seen from the side at more than about 15 km/h is smeared over more than its own length. ## Why the temperature reads low first A pixel only reaches the target’s full signal if the target stays in its view long enough. The response follows 1 − e^(−t/τ), where the dwell time t is roughly the target’s length along the direction of travel, minus the pixel footprint, divided by speed. Reading 90% of the real temperature rise needs about 2.3 τ (about 28 ms); 95% needs 3 τ (about 36 ms). Example: 13.5 mm lens, 10 m away (pixel footprint about 9 mm), share of the real temperature rise read along the 100 mm side / the 50 mm side of a target: - 3.6 km/h (1 m/s): 99.9% / 97% - 10.8 km/h: 92% / 68% - 18 km/h: 78% / 50% - 40 km/h: 50% / 27% - 80 km/h: 29% / 14% The 90% limit is about 12 km/h along a 100 mm side and about 5 km/h along a 50 mm side, and it barely changes between 5 and 20 m or between the 13.5 and 18 mm lenses. At road speed a hot spot still shows up, but its reading is pulled towards the background, often by half the real temperature rise or more. ## Looking ahead at an angle does not help Viewing obliquely shortens the smear in the image, but it compresses the target’s length along the direction of travel by the same factor. The time the target spends in one pixel’s line of sight is still its length along the travel direction divided by speed, whatever the angle. ## Turning counts too Rotation smears as well: angular rate × τ. A yaw or gimbal pan of 10°/s moves the scene by about 2.1 mrad during one time constant, about 2.3 pixels with the 13.5 mm lens (IFOV 0.9 mrad) and about 1.6 with the 9.1 mm lens (1.3 mrad) (calculated). ## What works - Drones: hover, or fly slowly (about 1 m/s) past the target you want to measure. - Vehicles: walking pace for small targets. Above that, use the image to find hot spots, then come back slowly to measure them. - Equipment that moves with the camera, mounted on the same vehicle, has no relative motion, so speed does not smear it. Vibration still matters. - A gimbal that tracks the target removes most of the relative motion. - Accurate readings of passing targets at road or rail speed generally need a cooled camera with a short integration time. That is a different class of product from an uncooled core. ## Distance: count pixels too For measurement, a common rule of thumb is at least 3 × 3 pixels on the target, so the maximum distance is target size ÷ (3 × IFOV). For a 50 mm target: - 9.1 mm (1.3 mrad): about 13 m in theory, but this fixed-focus lens is sharp only from 1.2 to 5.5 m - 13.5 mm (0.9 mrad, sharp from 3.8 m): about 18 m - 18 mm (0.7 mrad, sharp from 6.14 m): about 24 m With a stricter 5 × 5 rule, about 8, 11 and 14 m (all calculated). These figures come from the time constant and lens data in the module manual plus the model above. Check with your own target, speed and distance on a sample before you fix the design. [View MINI 640-2 Radiometric](https://thermal.torsuntech.com/products/mini-640-2-radiometric/) ## Tell us how the camera moves. Send the platform, its speed, the target and the distance. We reply within 24 hours on working days. [Discuss your requirements](https://thermal.torsuntech.com/contact/) --- TORSUN TECH · CATC Building 2, Shipai Town, Dongguan, Guangdong, China · info@torsuntech.com · Tel / WeChat / WhatsApp +86 180 2518 5996 · Enquiry: https://thermal.torsuntech.com/contact/