| 文章 | 年份 | 会议/引用量 | IF |
|---|---|---|---|
| Object tracking:A survery | 2006 | 7618 | |
| Object Tracking Methods:A Review | 2019 | 554 | |
| Multiple object tracking: A literature review | 2020 | 1294 | |
| Deep learning for multiple object tracking: a survey | 2019 | 145 | |
| Deep Learning for Visual Tracking:A Comprehensive Survey | 2021 | 432 | 23.60 |
| Deep learning in multi-object detection and tracking: state of the art | 2021 | 305 | |
| Deep Learning in Video Multi-Object Tracking: A Survey | 2020 | 807 | 6 |
others are coming soon...
-
定义：
It aims to infer the location of an arbitrary target in a video sequence, given only its location in the first frame -
应用：
traffic monitoring, robotics, autonomous vehicle tracking, medical diagnosis systems, activity recognition, and so on.
monitoring of traffic flow and detection of traffic accidentsASIMO humanoid robotpath-trackingtracking of ventricular wall and medical instruments controllearning activity patterns and human activity recognition（比如说VR)
- 挑战：
Illumination VariationBackground Clutters:the backgroundnear the targethas a similarcolor or textureas the targetLow ResolutionScale Variation:the ratio ofbounding boxesof the first frameand the currentframe is out ofthe rangeOcclusion:the target is partially or fully occluded(被遮挡)Change the target position:During themovement, thetarget may berotated,deformed, and soon.Fast Motion:the motion of theground truth islarge
- 方法：
feature-based, segmentation-based, estimation-based, and learning-based methods
generative methodsVSdiscriminative methods
都需要求 P ( Y ∣ X ) P(Y\mid X) P(Y∣X)，即已知样本x，求其属于类别y的概率。不同的是generative methods需根据公式P(Y∣X)= \\frac{P(X∣Y)P(Y)}{P(X)} 来求，但 ' d i s c r i m i n a t i v e m e t h o d s ' 直接求 来求，但\`discriminative methods\`直接求 来求，但'discriminativemethods'直接求P(Y\\mid X)。(Note that deep learning is belong to discriminative methods)
- 方法的评价：
RobustnessAdaptabilityReal-time processing of information
more details are provided in this paper：Object Tracking Methods:A Review