Video Surveillance and Tracking Methods

Video surveillance and tracking methods are concerned with automatically locating, following, and re-identifying objects or people across video frames and camera networks, turning raw pixel data into structured accounts of movement over time. The work draws heavily on deep learning — particularly convolutional neural networks — to solve problems like separating moving foreground objects from a changing background, maintaining consistent identities when targets are occluded, and matching a person who reappears on a different camera. These problems matter because reliable tracking underpins applications ranging from traffic monitoring to search-and-rescue, yet scaling to crowded scenes, adverse lighting, and real-time speed constraints remains genuinely difficult. Active research directions include building models that generalize across environments without exhaustive retraining, and developing tracking systems that remain accurate while respecting privacy and operating under computational limits imposed by edge hardware.

Works

82,451

Total citations

1,437,668

Keywords

Visual TrackingObject TrackingPerson Re-identificationBackground SubtractionConvolutional Neural NetworksReal-time Tracking