Deep learning X-ray

X-ray object recognition

X-ray security imaging is widely used to inspect luggage, mail, and vehicles to detect and discourage transport of illegal or dangerous items, such as firearms, explosives, etc. Currently, one line of development of X-ray security scanners is the adoption of multi-view systems that allow building a detailed 3-D structure of the scanned object. Another direction is automated recognition systems that aim for complete automation of scanning procedures eliminating humans from the process. With the increasing volumes of transported goods manual inspection of X-ray security imagery becoming a critical hurdle. Therefore, security agencies are pushing for significant technology’s step-ups in X-ray inspection capabilities. Benefits that automated systems would offer are obvious: significant cost reduction, increase in processing speed, improvement of consistency, and reduce opportunities for corruption.

deep learning x-ray gun detection report

There were a number of attempts to build a recognition system of illegal items in X-ray images using traditional computer vision approaches. However, due to a high variety and complexity of X-ray images, these attempts failed. Only with the evolution of artificial deep neural networks, the task became solvable. Deep neural networks allow working with x-ray images as accurately as humans do.

BitRefine offers a neural module that capable of recognizing illegal objects in x-ray images, not by material type/density, but directly by the object’s shape, exactly as the live operator does. Specialists at BitRefine have trained neural network modules with a large number of x-ray images of target objects, a firearm for example, until neural network learned to recognize it regardless of view angle, partial occlusions, etc. Recognition software not just recognizes the presence of an illegal item, but also shows the location of the revealed object by marking it with a bounding box. An important feature of BitRefine’s recognition solution that is checks detected objects several times as it moves through an X-ray scanner to ensure that found object is classified correctly.

x-ray object recognition knife detected

After an illegal item was recognized, the system sends an alert to the operator and highlight the object so that the operator can pay extra attention to this case. BitRefine’s recognition platform allows multiple ways of notifying staff, starting from a popup window or alarm sound to sending a command to external devices directly or via a network. In addition, all events are automatically saved in a database together with a timestamp and reference image of the detected item. Through the built-in reports tool, a supervisor can get a summary of all events, happened during a certain timespan.

An automated X-ray object recognition system is offered as an upgrade to existing screening points as well as for the newly build facilities. Main markets are airports, customs & border, ports, defense, critical infrastructure, and event security.

Related Solutions

We’ve got more video analytics to share. Let’s be friends. But you first...