目标跟踪相关综述文章

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…

1. 定义：
   It aims to infer the location of an arbitrary target in a video sequence, given only its location in the first frame

2. 应用：
   traffic monitoring, robotics, autonomous vehicle tracking, medical diagnosis systems, activity recognition, and so on.

• monitoring of traffic flow and detection of traffic accidents
• ASIMO humanoid robot
• path-tracking
• tracking of ventricular wall and medical instruments control
• learning activity patterns and human activity recognition（比如说VR)

3. 挑战：

• Illumination Variation
• Background Clutters:the backgroundnear the targethas a similarcolor or textureas the target
• Low Resolution
• Scale Variation:the ratio ofbounding boxesof the first frameand the currentframe is out ofthe range
• Occlusion: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

4. 方法：
   feature-based, segmentation-based, estimation-based, and learning-based methods
   

• generative methodsVS discriminative methods
  都需要求$P(Y\mid X)$，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)} $来求，但‘discriminativemethods‘直接求$P(Y\mid X)$。(Note that deep learning is belong to discriminative methods)

5. 方法的评价：

• Robustness
• Adaptability
• Real-time processing of information

more details are provided in this paper：Object Tracking Methods:A Review