Real Time Implementation of Drone Detectionusing TensorFlow and MobileNetV2-SSD

Drones are Unmanned Aerial Vehicles (UAV) that can be controlled by remote control or software methods. In recent years with the developing technology, drones have started to be used in all areas of our lives. The two most important factors that make drones attractive to use are low price and easy control. Today, UAVs, which provide convenience to people in many fields such as photography, video shooting, advertising, firefighting and search-rescue activities. Also, UAV is used by terrorist organizations in attack actions. Anti-drone systems are used to detect and locate such malicious drones. In this study, a prototype drone detection system called ATA-DRN is designed against drone threats. While creating ATA-DRN, 2 digital servo motors, Raspberry Pi 4 development board, high resolution Web Cam, 7inch HDMI LCD and power supply is used. With this designed system, drone approaching to the target is automatically detected with image processing methods without the need for manpower. In the study, first of all, a data set is obtained from pictures of different types of drones. The data set obtained is divided into two groups as train (80%) and test (20%) in order to train. The model to be used in real-time drone analysis is obtained using TensorFlow and MobileNetV2-SSD. Experiments are carried out by transferring the obtained model to the Raspberry Pi 4 development board. The accuracy of the model created as a result of experimental studies has been proven.