Mantis Vision

Mantis Vision's current public identity is a hyperspectral-imaging venture, not the older 3D-scanning business that appears in legacy company directories under the same name. Its website describes a camera and processing system that converts spectral measurements into processed data at video rates. The intended value is operational: hyperspectral imagery can expose material differences and spectral signatures that are not apparent in ordinary color imagery, while near-real-time processing could allow an operator or an onboard system to decide where to look or how to classify an object during a mission. The site also claims programmable selection of spectral bands, which, if demonstrated across applications, could reduce the need to design a completely different sensor for every mission. This is a systems proposition combining optics, spectral selection, edge processing, and application-specific analytics; the public site does not disclose enough architecture or specifications to determine which part is proprietary.

The most credible company-formation signal is the VTT Research LaunchPad page, which presents Mantis Vision as a science-based spin-off incubation team, names Egzon Delija as CEO and Antti Näsilä as CTO, and says the team is seeking seed funding. That places this record closer to a pre-commercial Finnish research spin-off than to the established 51-200-person, 2005-founded Petah Tikva 3D-imaging company found in older profiles under the same name. The current public materials do not reasonably confirm a legal founding year, registered headquarters, employee count, completed financing, recurring revenue, or named customers for the hyperspectral venture, so those legacy values should not be transferred to it.

The customer problem is clear but technically demanding. Defense and security users want to detect concealed, camouflaged, or otherwise hard-to-identify objects without waiting for an offline analysis pipeline. A Finnish government publication from the MATINE research program, dated March 2026, describes a 2024-2025 study involving Antti Näsilä that used a real-time near-infrared hyperspectral camera and an AI algorithm to detect a person in camouflage; it reports six analyzed hyperspectral cubes per second and materially better detection than a color-camera comparison. That is meaningful technical validation of the problem and approach, but it is not proof of a Mantis Vision product sale, military adoption, or a production-ready system. UAV integration is a sensible route because airborne platforms benefit from additional material information but impose tight constraints on mass, power, vibration, bandwidth, and latency. Earth observation and environmental or industrial remote sensing provide potential non-defense markets, but they also require validated calibration, repeatability, geospatial accuracy, and workflow integration. In each market, a camera demo is insufficient: buyers need evidence across illumination, weather, terrain, altitude, motion, and changing backgrounds, with a useful false-alarm profile.

Competitive pressure comes from established hyperspectral camera vendors, compact multispectral and snapshot systems, thermal and SWIR instruments, LiDAR, and increasingly capable RGB foundation-model pipelines. Mantis Vision's proposed edge is therefore not simply spectral resolution. It is the combination of real-time processed output, application-specific band selection, and integration into moving platforms. That combination could make a difference for time-sensitive detection, but the public record does not yet establish performance against comparable systems, sensor specifications, production readiness, or a defensible cost advantage. The company's approximately-one-kilometer camouflage-detection statement and Finnish Ministry of Defense testing reference are useful diligence leads, not independently verified proof of operational adoption or a government contract. They should be separated from the MATINE publication: the latter supports a government-funded research result, while the public record does not establish that Mantis Vision itself holds a contract or that the published experiment used a commercial Mantis-branded product.

Strategically, the company is relevant because it targets a sensing bottleneck in ISR and autonomous systems: distinguishing materials and objects when ordinary imagery is ambiguous. The opportunity is attractive for a small dual-use payload provider if it can secure a repeatable design win with UAV integrators or defense primes and then reuse the same hardware and algorithms in commercial remote sensing. The principal diligence questions are whether the VTT-originated intellectual property is owned or licensed on investable terms, whether the team can manufacture and calibrate sensors reliably, whether the system operates at genuinely useful frame rates and ranges, and whether early testing converts into paid pilots. Until those questions are answered, Mantis Vision should be considered a promising but unproven early-stage sensor company rather than a scaled defense supplier.