Chapter 14. Target Detection, Tracking, and Identification: ATR and Multi-Target Tracking

Target detection, tracking, and identification: ATR and multi-target tracking altitude of 25,000 feet. The sky beyond the canopy is bright blue. But beneath that calmness lurks a deadly tension. The radar warning receiver starts beeping. It's a sign that someone is watching me. Six green dots appeared on the cockpit's tactical display. You can't tell if it's an enemy, friend, or just a passing commercial aircraft. My heart is pounding. If one of these six dots is an enemy plane coming to kill me while flying at 1,000 kilometers per hour, I only have a few seconds left.

In the past, fighter pilots had to rely entirely on their eyes and senses for this moment. We called it 'Mk.1 Eyeball'. It means human eyes. Even if the radar detected something, the only way to know exactly what it was was to see it with your own eyes. The problem is that the enemy missile may have already been launched by then. The moment you see it with your own eyes, you die. This is the cruel reality of air combat. This is where ATR comes into play. ATR is short for ‘Automatic Target Recognition’.

To put it simply, it is a technology where the machine finds the enemy on its own and tells you “that’s an enemy” or “that’s an ally.” There is no need for people to stare at the radar screen with their eyes wide open. Artificial intelligence sees and judges on your behalf. At the core of the ATR system is a technology called deep learning. Deep learning is a method in which artificial intelligence learns patterns on its own by looking at millions of photos.

Just as a child can look at countless pictures of dogs and suddenly say, “That’s a dog!”, artificial intelligence also learns the shapes and characteristics of enemy tanks, enemy fighters, and enemy missile launchers. The shape of the turret, the angle of the wings, and the thermal pattern of the exhaust gases. I memorize these things over and over again. Let's take a look at how the ATR performs on the real battlefield. There is a device called Synthetic Aperture Radar (SAR). This radar can capture images of the ground through clouds, at night, and even when it is raining. However, the images sent by SAR are different from the photos we see.

It is a puzzling picture of gray and black mixed together. It would take even a trained analyst hours to decipher the picture. However, artificial intelligence provides an answer in 0.1 seconds. “Lidden in the shadows is a T-72 tank. The turret is rotated 15 degrees and the engine is on.” The US Department of Defense's 'Project Maven' showed the world the power of this technology. In the past, analysts had to sift through thousands of hours of video taken by reconnaissance drones.

Yes. If you stare at the screen until your eyes are glued to it, you may lose concentration and miss important things. However, artificial intelligence does not get tired. It analyzes videos 24 hours a day, selects only suspicious scenes, and shows them to people. This allows analysts to focus only on the decisions that really matter. This is why the F-35 Lightning II is called the ‘ruler of the sky.’ This fighter has an incredible ability called Sensor Fusion. Artificial intelligence combines all kinds of data coming from radar, infrared cameras, and electronic warfare equipment to create a single picture.

On the pilot's helmet display, enemy flags are displayed in a red box, and the aircraft type and armament status are also kindly written below it. “Su-35, estimated to be equipped with 4 air-to-air missiles.” Pilots no longer have to rack their brains trying to decipher radar scopes. But what if the enemy is not just one? This is where multi-target tracking comes into play. This means multi-target tracking. In modern warfare, the enemy does not come alone. Dozens of drones are swarming like bees, and a real fighter is hiding among them. Some are bait, some are real threats.

The human brain has difficulty keeping track of more than three or four objects at a time. When twenty dots appear on the screen, my mind goes blank. While you're trying to decide which one to deal with first, enemy missiles start flying. Artificial intelligence is different. Hundreds of targets can be tracked simultaneously. Assign each target a unique number, calculate its speed and direction, and predict where it will go. It's like a monster with a thousand eyes. In July 2024, the U.S. Air Force launched a project called 'ATA-AI (Advanced Tracking Architecture Using AI)'.

This is a project to develop next-generation target tracking technology with a budget of 99 million dollars, or over 130 billion won in Korean money. This is to deal with threats that are difficult to detect, such as stealth aircraft, hypersonic weapons, and drone swarms. Imagine a drone swarm. Hundreds of small drones swarm like bees. The human eye cannot tell which one is a suicide drone and which one is simply a disruptor. However, artificial intelligence analyzes the flight patterns of each drone. “Numbers 1 through 80 are simple decoys. They have no heat signature and only fly in a straight line.

But numbers 81, 95, and 112 are different. They are doing evasive maneuvers and have infrared signals. These three are the real threat.” Artificial intelligence performs this analysis in less than one second.

One of the core algorithms of this technology is the Kalman Filter. It is a mathematical formula developed in the 1960s that is used to predict the next location of a moving object. This is a time-honored technology that was also used in the navigation system of the Apollo spacecraft. Modern artificial intelligence goes one step further. By learning the enemy aircraft's past behavior patterns, it can even predict things like "This pilot likes to turn left" or "This type of aircraft tends to attack while lowering altitude." Lockheed Martin's Skunk Works is developing a missile avoidance system.

When an enemy missile comes in, it instantly knows which member of the squadron is targeting it and calculates the optimal evasive maneuver. In the past, pilots had to look at multiple displays to make decisions. Now, artificial intelligence compiles all that information and tells you, “Turn right to 5G right now!” China is not left behind in this competition. We are integrating a sophisticated electro-optical tracking system into the J-20 fighter jet and developing algorithms to detect US stealth aircraft, the F-22 and F-35. Attempts are also underway to neutralize stealth technology by combining quantum radar and artificial intelligence.

The European FCAS (Future Combat Aviation System) project pursues the concept of ‘Combat Cloud’. All aircraft on the battlefield are connected to one huge network, and each person shares what they see in real time. When a drone detects an enemy that a manned fighter cannot see, information is immediately transmitted. It's like hundreds of eyes connected to one brain. Ultimately, the laws of air combat have not changed. Whoever looks first, shoots first, and kills first wins. What has changed is speed. There are biological limits to human judgment speed.

It takes at least a few hundred milliseconds for the eyes to see it, the brain to recognize it, and the hand to move. However, the computational speed of artificial intelligence is in microseconds, not milliseconds. The human said "Huh?" Meanwhile, the artificial intelligence had already identified the target and opened the missile seeker. However, this does not mean that humans are no longer needed. No matter how many targets a machine can detect and classify, the final decision remains with humans. What if artificial intelligence mistakes a civilian airliner for an enemy plane?

We need someone who can say, "No, that's not the enemy." A hunter who wields a powerful hunting dog called artificial intelligence. That is the role of the future fighter pilot.