Advance and Refinement: The Evolution of UAV Detection and Classification Technologies

• Radar uses signals to locate UAVs and estimate velocity, but performance depends on UAV size, altitude, radar characteristics, weather, terrain, and reflective surroundings.Source: §1 · p. 2
• Radar clutter and radio interference can reduce detection range and target-location accuracy; higher-elevation installations help extend range by reducing ground-level obstructions [3].Source: §1 · p. 3
• RF sensing can track UAVs over long distances but requires sophisticated equipment, is interference-prone, and may estimate direction, range, and altitude inaccurately [3].Source: §1 · p. 3
• Optoelectronic systems offer precise target recognition but have shorter range than radar and depend on lighting and weather conditions.Source: §1 · p. 3
• Acoustic systems detect engine sounds at short range, but urban noise reduces reliability and smaller UAVs are harder to locate accurately.Source: §1 · p. 3
• Sensor fusion integrates diverse detection technologies to enhance UAV target accuracy and classification.Source: §1 · p. 3

2.1. The toolbar and its menus Technologies for detecting and recognizing UAVs based on radar systems

• Radar approaches include RCS-based statistical recognition, FMCW spectrogram and CNN classification [15], and 5G bistatic detection and localization using NLOS reflections and KNN.The system predicts presence and location on a 2x2 m grid and was tested indoors and outdoors; [15] uses STFT spectrograms and a ResNet-SP model.Source: §2.1 · pp. 5–6

2.2. UAV detection and recognition technologies based on RF sensors

• RF methods include passive sensing and compressed-signal learning, while XGBoost analyzes spectral signatures to detect UAV presence, type, and operating mode.Distributed passive-RF sensors use SDR and hardware-accelerated time-frequency processing; detection runs on embedded computers, with model identification possible at a fusion center.Source: §2.2 · pp. 6–7
• Passive RF sensing [22] achieved 99.94% precision at ranges up to 150m and 99.38% at ranges up to 350m; FEG and MC-DNN achieved 98.4% precision and 98.3% F1.The approach engineers discriminative RF features before a multi-channel network classifies UAV types and flight modes.Source: §2.2 · p. 7

2.3. Optical-electronic systems for UAVs detection and classification

• CNN-based UAV detection achieved 95% accuracy and outperformed SVM and KNN; YOLOv5 improved detection speed and precision over earlier YOLO versions, while modified YOLOv8 was used to detect small UAVs.Source: §2.3 · p. 9

2.4. UAV detection and classification technologies based on acoustic sensors

• Acoustic systems may be passive, active, or combined: passive systems can detect at long distances but cannot identify model, while active systems identify model but have limited range.Combined systems aim to provide both detection and recognition; acoustic approaches include signature-based deep learning and transformer models, and also analyzes operational anomalies using onboard microphone data.Source: §2.4 · p. 11

2.5. UAV detection and classification technologies based on Sensor Fusion

• In, infrared, video, and audio detectors achieved F1-measures of 0.76, 0.78, and 0.93, respectively, using machine-learning methods.The system combines sensor data to classify detected objects and track their trajectories.Source: §2.5 · p. 12
• RF-video fusion combines features from DNN and CNN processing streams and achieved 75% accuracy; radar-optical fusion merges coordinates and visual characteristics for target identification.The system fuses data at both the data and decision-making levels.Source: §2.5 · p. 12
• RF-acoustic fusion achieved about 91% accuracy at -10 dB SNR using LSTM features, while radar-optical fusion provided meter-level range and meter-per-second-level speed accuracy.The study reports improved performance in noisy, low-SNR environments; uses Fuse-Before-Track to remove uninteresting echoes and ground clutter.Source: §2.5 · p. 13

2.6. Use of UAV detection and classification technologies in modern Anti-Drone systems

• Anti-Drone systems fuse optical-electronic, acoustic, RF, and radar technologies, while AI supports UAV classification, recognition, sensor-data analysis, and decision-making.Cameras and infrared sensors provide visual detection; radar measures distance, speed, and direction; RF channels monitor drone signals, and acoustic sensors detect drone sounds.Source: §2.6 · p. 13
• Liteye AUDS combines radar acquisition, electro-optical tracking and classification, and directional RF jamming to protect infrastructure and support mobile missions.The system is also deployed on mobile and robotic vehicles for missions including convoy protection, electronic attack, wildfire response, and event security.Source: §2.6 · p. 14
• SkyWall Patrol uses an optical sight and compressed-air-launched net projectiles to capture UAVs, with parachutes on some projectiles reducing collateral-damage risk.It can operate standalone or integrate with wider security systems, and multiple units can protect a large strategic location.Source: §2.6 · p. 14
• 97% average classification accuracy; Kaspersky Anti-Drone reports detection range up to 5 km and RF range up to 1 km.It uses AI and neural networks to detect, classify, and track UAVs, drawing on sensors such as lidar, RF, radar, and PTZ cameras.Source: §2.6 · pp. 14–15
• Terrahawk Paladin uses radar, AI ballistic calculations, and day/thermal cameras to direct artillery against air targets at ranges up to 2 km, while ReDrone combines SIGINT and radar detection with radio and GNSS blocking.ReDrone is described as detecting, identifying, classifying, locating, and tracking multiple threats, then disrupting their communications and sending them off course.Source: §2.6 · p. 16
• The review identifies sensor-data fusion, AI and neural networks, and integration with existing security systems as development priorities, anticipating more efficient and accurate defenses.Described countermeasures include camera systems, detection radar, net guns, and cyber takeover systems.Source: §2.6 · p. 16

3. Conclusion

• Integrating radar, RF, acoustic, and optical systems is driving rapid advances in UAV detection and classification.Source: §3 · p. 17
• Machine learning and artificial intelligence play a pivotal role in refining technologies for more accurate and reliable systems.Source: §3 · p. 17
• Sensor fusion combines multiple data sources into holistic detection approaches that can adapt to varied UAV-monitoring challenges.Source: §3 · p. 17
• The review anticipates that better sensitivity and specificity, alongside AI-enabled real-time analysis and decision-making, will enhance airspace security and support safe civil and commercial UAV integration.Source: §3 · p. 17

Декларация о заинтересованности

• The declaration identifies financial interests or personal relationships that may be considered potential competing interests.Source: Декларация о заинтересованности · p. 21
• Ildar Kurmashev reports support from Manash Kozybayev North Kazakhstan University and funding from Kazakhstan’s Science Committee, Grant No. АР19679009.Source: Декларация о заинтересованности · p. 21
• The declaration states that any other authors have no known competing financial interests or personal relationships.Source: Декларация о заинтересованности · p. 21