Vehicle Detection using Images taken by Low-Altitude Unmanned Aerial Vehicles (UAVs)

Background/Objectives: This paper discusses the detection of vehicles in a dense area using low-altitude aerial images produced by unmanned aerial vehicles (UAVs). Vehicles in a dense area are difficult to detect accurately in images because of the narrow distance between parked vehicles. Methods/Statistical Analysis: This paper proposes a method that detects vehicles by applying a Histogram of Oriented Gradients (HOG) feature-extraction method to obtain information about vehicles found in images. Images used in the experiment were shot using a Phantom 3 Professional UAV developed by DJI Corp. Findings: Aerial images can be collected to measure traffic via a variety of methods. In recent years, studies on traffic volume using UAVs have been actively conducted. Until now, satellites, aircrafts, and helicopters have been used to obtain aerial images; however, the cost of such images is high, and these methods cannot respond to changes over time and weather in real time. As UAV technology has advanced in recent years, the methods of obtaining cost-effective and ultrahigh definition (UHD) aerial images have become available. The experimental results show the proposed approach can be used to detect vehicles that are densely packed in an area more effectively than the method using the conventional histogram of oriented gradients (HOG), which employs information about the brightness of images. Application/Improvements: In this paper, the detection of vehicles found in a densely packed area is performed using the HOG, which exhibits good adaptability to environmental changes.