Identify the platform.
Resolve the unit.
Hold the decision.
Modern conflict is observed through thousands of cameras at once. Vektra turns that feed into identification: which platform, which variant, which unit, with evidence a commander can review.
Recognition has become
a front-line capability.
Ukraine has shown what wide-area video, cheap airframes and electronic warfare do to a battlefield. Operator links fail. Navigation is jammed. Targets hide in treelines, under nets and behind decoys, and the volume of imagery exceeds what analysts can watch.
The systems that matter are the ones that keep recognizing when the link is gone, that tell a tank from a tractor and a launcher from a lorry, and that give a human a clear, reviewable basis for the decision.
From a platform class
to a configuration.
Each level is a model layer and a distinction your analysts care about. The crop tightens as the identification resolves.
- Platform classArmored fighting vehicle
Tracked hull, low profile, a turret: an armored vehicle, not a truck or a tractor.
- TypeInfantry fighting vehicle
Six road wheels a side, a small two-man turret with a long thin cannon, troop doors at the rear.
- VariantBMP-2
The 30 mm cannon and the turret shape separate it from the BMP-1 and from look-alikes on the same chassis.
- ConfigurationTurret traversed to the rear
Configuration details change what a platform is doing. Here the cannon points away from the guns.
From a frame
to an identification.
Each capability is a model layer built on the Vektra engine and shaped around the imagery, vocabulary and rules of engagement your organization defines.
Weapon system identification.
Resolve armor, artillery, rocket systems, air defense, radars, logistics vehicles, aircraft and naval platforms to type and variant. Distinguish look-alikes that share a chassis, and flag configuration details that change what a platform is.
Unit identification.
Read tactical markings, equipment mixes, vehicle counts and movement patterns as unit indicators. Build the order-of-battle picture from repeated observations rather than from a single frame.
Recognition under denial.
Run detection, tracking and identification on the airframe or vehicle itself. Continue when video and navigation links are jammed, and reconcile results with the common picture when the link returns.
Decoy and camouflage assessment.
Treat concealment as the adversary's design problem and ours. Fuse visible and thermal imagery, reason over shape and context, and report a decoy likelihood alongside the identification.
Small targets in large frames.
Find vehicles and positions that occupy a handful of pixels in satellite and high-altitude drone imagery, and cue higher-resolution collection for confirmation.
Human decision preserved.
Every identification carries its confidence, its source frames and its track history. Vektra supports the operator's and commander's decision. It does not replace it.
Scroll sideways for all six
The battlefield is the
hardest domain shift there is.
A detector trained on internet photographs has never seen a nadir view of a camouflaged howitzer in winter. Ours has, because we train it on yours.
Your vocabulary
Order-of-battle classes, variant lists, markings and priority targets become the model's taxonomy from the start, not an afterthought mapped onto generic labels.
Your sensors and conditions
Model layers are trained and evaluated on the altitudes, bands, seasons, terrain and compression you actually operate with, and re-evaluated on every release.
Your controls
Deployment on controlled infrastructure or on the edge, with local retention, air-gapped updates and audit of every identification, so the capability fits your doctrine and law.
We design for the cases
that break ordinary detectors.
Most recognition failures are predictable. They cluster around scale, shift, clutter, concealment, sensor physics and the link to the operator. We treat each as a design requirement.
Read the field noteTargets that are a few pixels wide.
Wide-area drone and satellite frames put most objects of interest under 32 pixels.
02 / DOMAIN SHIFTTrained in one world. Deployed in another.
Published benchmarks show detectors losing a large share of their accuracy when climate, terrain, season or sensor changes, with missed detections growing fastest.
03 / CLUTTER FALSE POSITIVESA shadow that reads as a vehicle.
Grid-based detectors reason locally, so shadows, rubble and scrap routinely trigger false alarms.
04 / CAMOUFLAGE & CONCEALMENTNets, treelines, dazzle and decoys.
Concealment is designed to defeat both eyes and models.
05 / THERMAL CONTRASTHot engine. Cold engine. Same target.
Infrared is strong against foliage and light fog, and weak when a target matches its background.
06 / DENIED LINKSWhen the operator feed goes dark.
Jamming severs video and navigation links at the worst moment.
Start with a defined
recognition problem.
A productive first conversation covers the platforms and units that matter, the imagery sources available, the conditions that have caused misses or false alarms, and where the recognition must run.
Vektra works under agreed handling, classification and access arrangements. Sample data and evaluation move through a controlled project channel after the initial scoping.
Discuss a defense programExport controls. Defense articles, technical data and defense services may be subject to the International Traffic in Arms Regulations (ITAR) and the Export Administration Regulations (EAR). We provide them only as authorized under U.S. export control laws, and we review end use and end user before any transaction. Do not send export-controlled technical data unless you are authorized to do so. Terms of use