Defense OS

Developed by Black Sage, Defense OS is a software platform that is aimed at orchestrating and allocating defense systems. It features excellent interoperability and supports passive and active sensor data fusion, threat prioritization, target classification, geofencing, scaling of assets around broad and diverse geographies, track deconfliction, accurate pointing of effectors and sensors, etc.

CLASSIFICATION OF TARGETS

Neural nets are the software appliances created to resemble a human brain in their ability to differentiate patterns. Neural nets are widely used in Defense OS as they can easily recognize patterns and rapidly determine which class the target belongs to.

VALIDATION OF CLASSIFICATION PERFORMANCE

Cross-validation is a way to evaluate the performance of the target classifier. Once classification information is updated in the system, a new classifier and five-fold 80/20 sampling cross-validation appear. Every validation step returns both false and true positive rates and promotes a receiver functioning feature curve that indicates actual classification performance. The data received during these steps are averaged and presented to the operator in the dashboard of Defense OS.

SELECTION OF FLEXIBLE ATTRIBUTES

A plethora of target attributes is currently available, including modern sensors and the ones that describe objects: height, width, longitude, latitude, acceleration, velocity, surface area, heading, displacement, and other features. Defined as a multidimensional signature of an object, they are used mainly by the classifier to achieve proper pattern recognition. With the diversity of target attributes offered, operators can easily solve apparently intractable problems with false alarm by selecting the ones that are included to control the behavior and performance of the classifier.

BIRD and UAS

Many problems appear during the process of airspace monitoring, with distinguishing between UAS and birds being one of the most common and challenging ones. However, Al recognition of heading, size, lon, lat, acceleration and velocity patterns simplifies the task and contributes to its reliability and automaticity.

PERSON and VEHICLE

Deterministic alarms usually confuse people and vehicles. Al can differentiate unique peculiarities, such as the times or paths of travel to precisely classify when deterministic alarms cause false positives.

ROTORCRAFT and FIXED

Deterministic systems have problems distinguishing fixed-wing aircraft and rotorcraft, as they share some flight specifications. Al can cope with this task easily by learning from the surface area and altitude deltas.

UNLIMITED LEARNING

The classifier does not depend on attributes, classes or sensors. Similar to the human brain, it may be trained on literally any type of input information and will learn and modify the behavior accordingly.