A Vision-Based Amateur Drone Detection Algorithm for Public Safety Applications

Drones will be widely used in the smart cities of the future for the wireless approach, supplying goods, and for conserving the security of smart cities. Apart from the various benefits of the drones, they pose significant challenges and public concerns that need to be tackled. We propose an amenable framework to detect malicious drone and ensure public safety. The proposed model is validated using 400 images with various challenges such as occlusion, scale variation, blurriness, background clutter, and low illumination. After extracting the features using ResNet50 model, we applied SVM's RBF kernel for the classification. The classification accuracy presented in the confusion matrix for this model came out to be the highest in comparison to other state-of-the-art models.