Object Tracking in Videos Involves Estimating the State of Target Objects from Previous Information

EasyChair Preprint no. 12519

Abstract

Object tracking in videos is a crucial task in computer vision that involves estimating the state of target objects based on previous information. It plays a significant role in various applications, such as surveillance systems, autonomous vehicles, human-computer interaction, and video editing.

The primary objective of object tracking is to locate and follow objects of interest as they move within a video sequence. This process typically involves three main stages: initialization, detection, and tracking. In the initialization stage, the target object is identified and its initial state is estimated. This can be achieved through manual annotation, user interaction, or automated techniques such as background subtraction or object detectors.

Once the object is initialized, the detection stage aims to locate the target object in subsequent frames. This is typically done by employing visual features, such as color, texture, shape, or motion information, to discriminate the object from the background or other objects in the scene. Various algorithms, including template matching, correlation filters, and deep learning-based methods, are commonly used for object detection.

After the object is detected, the tracking stage involves estimating the state of the target object over time. This includes estimating its position, size, orientation, and other relevant attributes. The state estimation can be achieved through various techniques, such as Kalman filters, particle filters, graph-based methods, or deep learning-based approaches. These methods leverage the temporal coherence of the video sequence and exploit the spatio-temporal information to accurately track the object.

Keyphrases: Artificial, intelligence, vision

@Booklet{EasyChair:12519,
  author = {Favour Olaoye and Kaledio Potter},
  title = {Object Tracking in Videos Involves Estimating the State of Target Objects from Previous Information},
  howpublished = {EasyChair Preprint no. 12519},

  year = {EasyChair, 2024}}