Uncooled Targeting System for Counter-UAS Solutions
Discover how SpearX is revolutionizing uncooled targeting systems for counter-UAS applications. Learn about the challenges of cold start and innovative search architectures that enhance time-to-lock metrics, ensuring effective tracking even in difficult conditions.
10/7/20262 min read
We call it "COLD START".
1. No model, no prediction The Kalman stack works only after the first good point. Before that, every choice rests on one raw frame. You cant speed up cold start with the same filter.
2. Wide FOV searches, narrow FOV sees To find a target you need a wide field. To lock it, you need pixels. At 0.5 mrad IFOV, a Shahed at 2 km is 2-3 pixels: a dot with no shape. Shape is what Part 2 used to reject flares. On cold start, shape is off.
3. Scanning the whole frame is a trap First instinct: run a heavy detector over everything. A 640×480 frame is ~307k pixels; a 100×100 ROI is 10k. Thats 30× the work, straight into the latency we turned into meters in Part 1.
4. Cueing shrinks the window by orders The answer isnt look harder, its look where you were told. Radar/RF gives a direction. A full wide FOV is hundreds of mrad; a cue window is a few. Search area falls by thousands.
5. Coarse → fine is a pipeline Two layers on different clocks. Coarse: fast, cheap, low threshold, must not miss. Fine: heavy work only in the narrowed ROI, where shape turns on. While fine picks the first candidate, coarse hunts the next.
6. Miss cost vs false-lock cost With no history, every bright pixel is a candidate. High threshold at long range means missing a real one. A missed Shahed at 600 km/h may not return. Threshold follows miss cost, not a pretty image.
7. The metric is time-to-lock, not range Specs love detection range. For CUAS, the real number is time from first appearance to stable track. At 350 m/s, a 500 ms lock is 175 m the target flew while you found it.
Bottom line:
Cold start is solved by search architecture, not a better sensor:shape off at first;cueing shrinks the window;coarse → fine is a pipeline;metric is time-to-lock, turning into meters.
Prev. Parts 1-2 lived with a born track. Part 3 is how to birth it from noise. Part 4: the target vanishes - cloud, smoke, lost contrast.
Why lost visual isn't lost target. How is your cold start organized? What hurts more - a miss or a false lock?
Over the past year, SpearX has been developing an uncooled (LWIR) targeting system for counter-UAS (CUAS) systems.
Part 1 - showed why we cant just raise fps. Part 2 - showed how to hold a track when a flare lies.
Both assumed the track existed. At the start it doesnt. No model, no history. The most fragile moment is the first lock.
Related : GEOCOM Co. LLC www.geocomco.eu DeepTechRnD www.deeptechrnd.eu SpearXAgro www.spearxagro.eu
Contacts
SpearX ￪ Strike with Precision
+359 877 620 210
D-U-N-S® 525540791
© 2025 - 2026. All rights reserved. SpearX Projects. Geocom Co. LLC
SpearX is an EU-Ukrainian company located in a strategic areas for R&D in UAV