Challenging the Implementation of Video for ATC

Video offers many advantages that advanced tracking systems could exploit to reach a higher level of performance. Yet, performing real-time video object detection and tracking in the natural outdoor environment is a difficult task, belonging to the general class of multiple-target tracking problems. Multiple-target tracking systems are being developed to address a large number of applications, including Air Traffic Control (ATC). In implementing a functional ground ATC system reporting good quality tracks from video, one has to overcome difficulties related to dynamic background and lighting changes. These include the visual impact of swaying trees moving in the wind, as well as the changing position and intensity of cast shadows according to the position of the Sun and cloud conditions in the sky. Continuously tracking an outdoor scene involves adapting from daytime conditions to nighttime conditions and reducing as much as possible the negative impact of fog, rain or snow on the visual quality of the images.

Hence, the first and foremost challenge in video multiple-target tracking in the outdoors is to process noisy raw data (aka images) to form consistent observations from frame to frame in real-time. The reliability of this initial stage of the tracker will strongly influence system performance figures such as detection and false alarm rates. For the sake of simplicity, we will limit the discussion to a single fixed camera providing images at a constant frame rate. Having access to a full network of cameras does provide additional perspectives to the system without affecting the generality of camera-based object detection.

To download the full Feature Article, click here.