Detection of Vehicles in Aerial Images

Abstract

Vehicle detection using aerial images uses images taken from a great distance, usually using satellites, to detect different types of vehicles on the ground. These images are processed using machine learning tools and techniques to distinguish the images from the background. A great challenge that is faced in the task is to overcome the difficulty in feature extraction and detection that is introduced due to the height and the varying environment on the ground. In our solution, we aim to improve the accuracy of vehicle detection in aerial images while trying to increase the efficiency with which the system produces the results. Vehicle detection is of great importance in traffic monitoring and control, urban planning, rescue tasks, criminal surveillance and military activities. In the modern days, urban planning heavily relies on using the traffic patterns to predict traffic flows and the density of the traffic at different times and locations for an effective management of traffic and the projection of traffic patterns in the future for planning and development purposes. Security surveillance systems also make use of vehicle detectors to detect any intrusion. We propose a system that allows a user to upload images and view all the vehicles present in those images. The system also allows the user to feed real-time images into the system to detect the location of vehicles on live data.