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.
