-10CNN-Based UAV Detection and Classification Using Sensor Fusion

This paper proposes a detection and classification method for unmanned aerial vehicles, commonly called drones, using sensor fusion schemes. Datasets drone are collected by field measurements of actual drones the optical camera, radar, audio microphone as well obtained from open online databases. In first stage proposed method, conducted convolutional neural network (CNN) models separately trained images, radar range-Doppler maps, spectrograms. Then, CNN output probabilities combined multinomial logistic regression to improve surveillance accuracy through optical, sensors. Numerical simulations performed with experimental data datasets. From results, it is verified that can up 15.6% enhance 28.1% in terms F-score, compared individual sensing