Increasing protection with drone detection analytics

Guest authors

Dr. Markus Schoeler, Lead Engineer of Video Analytics at Dedrone, explores the new ways of dealing with drone intrusion.

In January, the U.S. Department of Transportation and Federal Aviation Administration officially reached one million registered drones in US airspace. Registration programs are emerging throughout the world. In its 2017 forecast, the FAA estimated that hobbyist drones will more than triple in size over the next 5 years, to over 3.5 million units by 2021.

Why is drone detection necessary?

 Fences alone are no longer enough to prevent intruders from placing dangerous devices within reach of people and critical infrastructure. This includes places such as industrial facilities, data centers, stadiums, correctional institutions, airports, and military installations, just to name a few.

Additionally, just as the internet opened up a new way for hackers to harm organizations, drones have enabled similar situations such as corporate espionage, smuggling, terrorism, and hacking. As a result, drones now represent a serious new security threat to any organization.

The risk of disrupting control systems or doing physical harm is a danger that companies are trying to solve by integrating cameras with applications to provide a complete drone detection solution.

How does the technology work in practice?

  1. Cameras with in-built capabilities can analyze the video stream.
  2. The program learns the scene, makes note of obstacles and movement patterns, such as aircraft trajectories.
  3. The software can then classify all objects into a database which is able to differentiate between drones, birds, helicopters and passenger aircrafts.
  4. The video feed is analyzed frame by frame to identify if any object matches the specific characteristics of a flying drone.
  5. Once a drone has been detected, the software begins automatically recording for future evidence and alerts security personnel.

As soon as cameras are connected, a software can begin to analyze the video stream. The software learns from the scene and makes a note of all obstacles, such as trees, and movement patterns, like a local highway or aircraft trajectories. The technology understands the setting, maintains a model of the scene at all times and detects any abnormal activity.

The software is able to classify all objects using a database that encodes visual cues, like shapes, colors and sizes to determine what something is. Video feed is analyzed on a frame by frame basis, potential objects are extracted, tracked, and matched up with the database to automatically identify drones.

The visual cues as well as the way it moves can tell drones apart from other objects. A drone has very specific characteristics that databases can analyze and compare with images of other objects detected through cameras, such as birds, helicopters, passenger aircraft, cars, and trains. At each update, a database can process millions of different images of drones, aircraft, birds and other objects.

Once the software classifies an object as a drone, it begins automatically recording and ensures the clearest line of sight with the drone and its flightpath. This all happens in the blink of an eye.

What’s the benefit of incorporating video into a drone detection installation?

With the visual information obtained through cameras, security personnel have a greater chance of establishing the reasons behind the drone’s intrusion.  Video shows security personnel physical evidence that will enable them to determine if the drone is carrying a sensitive or illegal payload, contraband, has attached cameras for spying, or hacking tools to manipulate networks.

To show an unimpeded skyline and make it easier for sensors to capture accurate data, cameras can be positioned above trees, hills, or other visual distractions. To this end, cameras can be placed throughout a customer’s property to create an accurate horizon, further customizing the installation process.

Drone technology is quickly evolving and changing. Video analytics can classify any drone, including homemade and prototypes. This is especially important as many new and emerging technologies may not use common radio frequency (RF)/WiFi communication protocols. Pilots who are looking to avoid detection, like those dropping contraband into a prison, frequently turn off their remote control and use a GPS-guided drone to follow a pre-defined path. If there are no radio waves emitted, then video will be the only way to detect an intruder.

Another benefit is that video can also detect silent or quiet drones, as well as multiple drones at a single time (such as if they are flown in swarms).

Drones can be detected in a variety of ways, yet having video footage provides its own advantages. In addition, they can be detected by other passive sensors, like RF/WiFi, which detect the communication protocol, or connection between the pilot’s controller and the drone. Microphones can also be used to pick up on the audio signals that only a drone can make.

Each technology has its pros and cons, and a combination of different sensors provides the best coverage against possible drone intrusions.

Will distance and low light affect detection?

The distance at which video analytics can detect a drone largely depends on the type of camera used. A camera’s field of vision depends on the opening angle of the lens. Some cameras have wide angles, and some are narrow. Like the illustration shows below, the wider the field of view, the lesser the distance a camera is able to capture drones at in sufficient resolution. Some can see activity from hundreds of yards away from where they are installed. Various software can work with all cameras and analytics to aggregate information on to a protected site, ensuring complete coverage of the property.

Detection in low light conditions will come down to the type of camera, its capability and deployment, such as if the camera includes night vision technology or if the area is well lit. Most drone pilots avoid flying at night, and many are unable to fly in inclement weather, as it may damage the hardware or prevent safe flying. However, illuminating the area is not always possible and some criminals will want to fly at night to avoid detection, which is why video should be one element but not the sole element, of a detection ecosystem. The more types of sensors deployed, the better the detection.

Drone detection is constantly evolving

Our video analytics software is advancing every day and learning from customer detections providing more data. Like an antivirus software, DroneDNA and the core of our video processing pipeline is constantly evolving. DroneTracker is where artificial intelligence and drone detection meet, and we’re using the most up-to-date information and state-of-the-art technology to ensure each of our customers can protect their airspace from all drone threats.