An X-ray image reveals what is inside an object; reading that image correctly is a distinct skill. XRayDetect layers an automatic detection and object recognition stage onto the dual-energy image-processing pipeline: classical computer vision paired with machine learning and deep learning. From operator decision support all the way to fully automatic decisions — for security screening, food safety and non-destructive testing (NDT).
The same groundwork — identifying objects and materials inside an X-ray image — carries over to three different industries: security screening, food safety and non-destructive testing. Each domain is then tailored with its own data and its own rules.
We build the AI on top of our established image-processing know-how. Classical computer vision (segmentation, feature extraction) gives a dependable, explainable baseline, while deep learning handles the intricate patterns where classical methods fall short — blended in whatever proportion the task calls for, from decision support to fully automatic mode.
The same recognition core, tailored per domain with its own data and rules — from operator decision support to fully automatic verdicts.
In conveyor X-ray scanners we give the operator decision support: weapons, knives and dangerous or prohibited items are flagged automatically. Drawing on material data from dual energy (organic / inorganic / metal), we highlight suspect regions and cut down on human error.
In-line X-ray inspection for food safety: spotting foreign bodies like metal, glass, stone, bone and dense plastic, along with fill / missing-piece and package-integrity checks. X-ray looks straight through the packaging and the product where optical systems stop at the surface.
Automatic detection and classification of defects, voids, cracks and assembly errors in industrial parts. Consistent, repeatable verdicts in series production — quicker and more objective than inspecting by hand.
Segmentation and feature analysis give an explainable baseline, combined with CNN-based detection / classification for the intricate patterns. The two are blended in whatever proportion the task calls for.
Models run at production-line or conveyor speed, with no latency — a natural fit into the existing dual-energy image-processing pipeline and the operator interface.
Data stays on site, with no reliance on the cloud. The false-alarm versus miss trade-off is tuned to the task, with retraining on fresh samples and ongoing performance monitoring.
We build the AI on top of our established image-processing know-how. Classical computer vision (segmentation, feature extraction) gives a dependable, explainable baseline; deep learning handles the intricate patterns where classical methods fall short. We blend the two in whatever proportion the task calls for.
The recognition layer never operates on its own; it joins with our other competencies to form an end-to-end solution:
Models, code and documentation with the rights to use and modify — no lock-in to the contractor.
Data collection, labelling, training and independent validation on the objects, foreign bodies or defects you need to find.
Real-time inference at conveyor speed, deployed on-prem / edge with data kept on site — no cloud dependency.
Tell us which objects or defects you need to find and which equipment you run — we'll assess feasibility and propose a pilot together with a roadmap. Where you can, prepare some test samples. Reach us at info@xraydetect.com.