Flying objects Classification Using Trajectory Images and Convolutional Neural Network

In this paper, we present a new method for flying object classification and recognition based on trajectory images and deep learning approach. First, the video sequence is passed through a Gaussian mixture model (GMM) for flying object detection and tracking. Then the trajectories are determined by the different positions of the objects at successive frames. After that, we collect a large database of trajectories of two classes of objects: drones and birds. Then, the convolutional neural network (CNN) is applied to classify these objects into two classes. The LeNet5 is one of the best choice used in image classification. The CNN is trained using 80% of the database and the rest is used for testing. Experimental results are conducted to demonstrate the performance of the proposed method for drones and birds objects classification. In addition, the automatic approach helps and facilitate to military services to recognize drone from others objects.