small object

pedestrian detection

visual tracking

face detection

tracking

video quality

camouflaged object

object detection

moving object

Displays

Multiple Object Tracking (MOT) has emerged as a hot issue in the field of computer vision recently. MOT has academic and commercial potential in urban public security, military, self-driving and other fields. Although researchers have put forward different approaches to solve the problem, it is still challenging due to factors such as appearance mutation and severe object occlusion. This study makes a comprehensive research on the problem of detection-based video multi-object tracking. The main contributions of this review are as follows: 1) The important aspects of MOT system are studied, including the basic principles, main framework, technical difficulties, classification and evaluation; 2) This work review the classification and latest progress in object detection due to its close relevance to the field of MOT; 3) This work introduce a taxonomy based on the MOT problem on this basis, the original and improved versions of each classical tracking algorithm are discussed, analyzed and compared; 4) This work summarize the results of 40 algorithms on popular datasets to provide quantitative comparison and analysis of experimental results. Finally, the possible research directions for the future are put forward, by analyzing the commonness of several methods with excellent performance. • The important aspects of Multi-object tracking system are studied, including the basic principles, main framework, technical difficulties, classification and evaluation. • We review the classification and latest progress of object detection due to its close relevance to the field of MOT. • We introduce a taxonomy based on the MOT problem on this basis, the original and improved versions of each classical tracking algorithms are discussed, analyzed and compared. • We summarize the results of 40 algorithms on popular datasets to provide quantitative comparison and analysis of experimental results. This paper finds the commonalities of the methods with excellent performance and puts forward the possible research direction of MOT in th...

A survey of detection-based video multi-object tracking