.png){kind=icon}
Return
CNA-DeepSORT algorithm for multi-target tracking
.png){kind=icon}
PRE
AI
DOI：10.1007/s11042-023-15813-z.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In recent years, multi-target tracking algorithms have been developed rapidly. However, in multi-target tracking, mutual occlusion and cross between targets and sudden disappearance and reappearance of targets in videos can easily occur, which could only result in missed detection, false detection, and wrong ID switching. To address the above problems, the CenterNet attention DeepSORT algorithm (CNA-DeepSORT) proposed in this paper incorporates a CenterNet network with channel attention mechanism in the original detection part of the DeepSORT algorithm instead of Faster R-CNN, and designs a multi-scale feature extraction module with the DeepSORT algorithm in the multi-scale feature extraction module and designed a pedestrian recognition network combined with the DeepSORT algorithm. These improvements lead to a 3.7% improvement in MOTA metric, 1.6% improvement in MOTP metric, 238 fewer false ID switches, 2627 fewer FP metrics, 3943 fewer FN metrics, a decrease in run speed, and a 4 Hz reduction in frame rate compared to the original DeepSORT algorithm. improved by 3. 7, and there is some improvement in handling the occlusion problem of multi-target tracking, and the false and missed detection of targets during ID switching is reduced.
Keywords:
Multi-target tracking
DeepSORT
Target detection
Attention mechanism
Multi-scale
Journal
IF:
3
Papers:
2.0W
Citations:
3.2W
Organization
Cited Papers
ShareSaveShareSaveShareSaveRecommendation system using feature extraction and pattern recognition in clinical care systems
PREAI
ShareSaveDamage in the Multiethnic Malaysian Systemic Lupus Erythematosus (SLE) Cohort: Comparison with Other Cohorts Worldwide.png)
PLOS ONE
OAAI
PLOS ONEIF0
OAAIShareSaveTime Series Analysis and Forecasting of Air Pollution Particulate Matter (PM2.5): An SARIMA and Factor Analysis ApproachIEEE ACCESS
OAAI
IEEE ACCESSIF3.6
OAAIShareSaveShareSave
View more