Video Surveillance and Tracking Methods

Video surveillance and tracking methods are concerned with automatically locating and following objects or people across video frames, using techniques that range from classical background subtraction and foreground segmentation to modern deep learning architectures such as convolutional neural networks. The practical stakes are high: reliable real-time tracking underpins applications in public safety, autonomous navigation, and crowd analysis, where errors in detection or identity association carry direct consequences. A persistent challenge is person re-identification — recognizing the same individual across different cameras, viewpoints, or lighting conditions — where appearance alone is often insufficient and models must learn robust, generalizable representations. Active research is pushing toward tracking systems that remain accurate under occlusion, crowding, and domain shift, while also grappling with the computational demands of processing high-resolution video streams in real time.

Works

83,679

Total citations

1,460,243

Keywords

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