原文链接:https://mp.weixin.qq.com/s/nLPfErOciUnxhzXx9aI-yA
论文阅读:[1] Z. Wang, C. Jin, L. Sun, and Y. Li, “WRDI: A Multimodal Dataset of mmWave Radar Data and Image,” in 2021 IEEE International Conference on Big Data (Big Data), 2021, pp. 3209–3214. doi: 10.1109/BigData52589.2021.9671545.
Abstract
-
mmWave radar and image data fusion 在 Security monitoring field 作用很大
- real-time
- high-sensitivity
- 全天候全天时
- effectively filter out fake objects
-
挑战 :lack of a comprehensive mmWave radar and image multimodal dataset
- 已成为 a bottleneck in developing robust object detection and tracking models
-
本文贡献: WRDI
- WRDI: a mmWave radar data and image multimodal dataset with 5000 frames
- 意义 :the first comprehensive mmWave radar and image multimodal dataset for security monitoring field.
1 Introduction
安防( intrusion events )领域:mmWave Radar and Camera can complement each other
-
Camera
- main terminal device
- 优点:capture fine texture and rich color information of the object
- 缺点:difficult to receive enough information to under bad condition
-
mmWave Radar
- 优点: wide sensing range + strong obstacle penetration ability, 不受 weathers and lighting conditions 影响
- 缺点:sparse and lack rich scene information
-
⇒ \Rightarrow ⇒ fuse mmWave radar and optical sensor data in security monitoring system
数据融合 (camera + mmWave radar) 研究现状
-
现有研究
- [1] radar + lidar + camera: detection and tracking
- [2] fuse in feature extraction
- [3] radar + image
- [4] radar + image
- [5] radar + image
-
缺陷
- 不适用 :most of the radar and image multimodal datasets are collected in urban traffic roads