projectRecord
0. Tools
.1. pymmw
- chrome-extension://ikhdkkncnoglghljlkmcimlnlhkeamad/pdf-viewer/web/viewer.html?file=https%3A%2F%2Fcore.ac.uk%2Fdownload%2Fpdf%2F225625415.pdf%23page%3D6#page=6
- 学习这里的每一个特征
Constapel, Manfred, Marco Cimdins, and Horst Hellbrück. “A Practical Toolbox for Getting Started with mmWave FMCW Radar Sensors.” 4th KuVS/GI Expert Talk on Localization (2019). [pdf] [code]
a toolbox composed of Python scripts to interact with
TI's evaluation module (BoosterPack) for IWRxx43 mmWave sensing devices.
- access to particular OOB firmware versions
- access to particular OOB firmware versions
- 2D plots
range and noise profile
Doppler-range FFT heat map
azimuth-range FFT heat map
FFT of IF signals
- 3D plots
CFAR detected objects (point cloud)
simple CFAR clustering
- Data capture
range and noise profile with CFAR detected objects
.2. AWR1843-Read-Data-Python-MMWAVE-SDK-3-
Python program to
read and plot the data
in real time from the AWR1843 mmWave radar board (Texas Instruments, MMWAVE SDK 3). The program has been tested withWindows and Raspberry Pi
and is based on theMatlab demo from Texas Instruments
.First, the program configures the Serial ports and sends the CLI commands defined in the configuration file to the radar. Next, the data comming from the radar is parsed to extract the 3D position and doppler velocity of the reflected points. Finally, the 2D position of the reflected points is shown in a scatter plot.
.3. pyroSAR
The pyroSAR package aims at providing a complete solution for the
scalable organization and processing of SAR satellite data
:
- Reading of data from various past and present satellite missions
- Handling of acquisition metadata
- User-friendly access to processing utilities in SNAP and GAMMA Remote Sensing software
- Formatting of the preprocessed data for further analysis
- Export to Data Cube solutions
.4. OpenRadar
- Reading raw ADC data.
- Preprocessing data in DSP stack.
- Utilizing preprocessed data for tracking, clustering and machine learning.
- Different demo implementations from TI and our own explorations.
- openradar.readthedocs.io
.5. mmWaveRadarPro
1. Object Detection
.1.CameraRadar Fusion
Nobis, Felix, et al. “A deep learning-based radar and camera sensor fusion architecture for object detection.” 2019 Sensor Data Fusion: Trends, Solutions, Applications (SDF). IEEE, 2019. [pdf]
- provides a neural network for object detection based on camera and radar data.
.2. CenterFusion
Nabati, Ramin, and Hairong Qi. “Centerfusion: Center-based radar and camera fusion for 3d object detection.” Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. 2021. [pdf]
- first uses a
center point detection network
to detect objects byidentifying their center points on the image
- solves the
key data association problem
usinga novel frustum-based method
to associate theradar detections
to their correspondingobject's center point
- The
associated radar detections
are used to generateradar-based feature maps
to complement theimage features
, andregress to object properties such as depth, rotation and velocity
.
.3. RODNet
Wang, Yizhou, et al. “Rodnet: Radar object detection using cross-modal supervision.” Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. 2021.[Paper] [Dataset] [code]
.4. Radar-RGB-Attentive-Multimodal-Object-Detection
Yadav, Ritu, Axel Vierling, and Karsten Berns. “Radar+ RGB Fusion For Robust Object Detection In Autonomous Vehicle.” 2020 IEEE International Conference on Image Processing (ICIP). IEEE, 2020. [pdf] [code]
.5. radar-ml
Radar-based
recognition and localization of people and things
in the home environment has certain advantages over computer vision, including increaseduser privacy
,low power consumption
,zero-light operation
andmore sensor flexible placement
.
- accurately
detect people, pets and objects
using low-power millimeter-wave radar- self-supervised learning, leveraging conventional camera-based object detection, can be used to generate radar-based detection models.
- Arrange a
camera
and aradar sensor
to share acommon view of the environment
.- Run the radar and a camera-based object detection system to
gather information about targets
in the environment.- Create
ground truth observations
from the radar when it senses targets at the same point in space as the object detector.- Train a machine learning model such as a Support-Vector Machine (SVM) or a deep neural network (DNN) on these observations and check that it has acceptable accuracy for your application.
- Use the trained machine learning model to predict a novel radar target’s identity.
.6. RAI Image based ship detection
.7. PersonCount
Wen, Longyin, et al. “Detection, Tracking, and Counting Meets Drones in Crowds: A Benchmark.” arXiv preprint arXiv:2105.02440 (2021).论文链接 项目链接
通过提出 STNNet 方法来共同解决无人机拍摄的拥挤场景中的密度图估计、定位和跟踪。值得注意的是,作者设计了相邻上下文损失来捕捉连续帧中相邻目标之间的关系,对定位和跟踪是有效的。为了更好地评估无人机的性能,作者还收集并标记了一个新的数据集:DroneCrowd。,其中包含了用于密度图估计、人群定位和无人机跟踪的头部标注轨迹。并称希望这个数据集和提出的方法能够促进无人机上
人群定位、跟踪和计数
的研究和发展。
.8. milliEye
Xian Shuai, Yulin Shen, Yi Tang, Shuyao Shi, Luping Ji, and Guoliang Xing, milliEye: A Lightweight mmWave Radar and Camera Fusion System for Robust Object Detection. In Proceedings of Internet of Things Design and Implementation (IoTDI’21) [imaging]
2. RainFow Prediction
.1. CIKM-Cup-2017
- 赛题提供10,000组的雷达图像样本。每组样本包含60幅图像,为过去90分钟内(间隔6 min,共15帧),分布在4个高度(0.5km, 1.5km, 2.5km, 3.5km)上的雷达反射率图像。
- 每张雷达图像大小为[101,101],对应的空间覆盖范围为101×101km。
每个网格点记录的是雷达反射率因子值Z
。反射率因子,表征气象目标对雷达波后向散射能力的强弱,散射强度一定程度上反映了气象目标内部降水粒子的尺度和数密度
,进而推测其与降水量之间的联系。
- 目标
:利用各个雷达站点在不同高度上的雷达历史图像序列
,预测图像中心位于[50,50]坐标位置的目标站点未来1-2小时之间的地面总降水量
,损失函数为降水量预测值与真实值的均方误差
。
-
图像拼接: 样本与样本之间并不完全独立,图像样本之间存在一定的重叠,可以通过模板匹配的方式寻找样本之间的坐标关联特性。通过样本之间的局部图像拼接,能够将一系列小范围的局部雷达图像恢复到空间更大范围的雷达图像,进而获得关于云团更加整体的特性。
-
**轨迹追踪:**泰勒冻结假设(Taylor Frozen Hypothesis),流场中存在显著的时空关联特性,即可以认为雷达反射图中云团在短时间内趋向于在空间以当地平均对流速度平移,短时间内并不会发生外形或者反射强度的剧烈改变。即监测点x处在未来τ时刻后的雷达信号f,能够通过平均对流速度U
-
**特征提取:**特征包含
时间外插反射率图像
,时间空间的矢量
,云团形状
的统计描述三部分。
.2. RadarGUI
PyQt 5 CINRAD雷达基数据处理可视化软件。 基于Python 3.6,用于雷达回波绘制和显示,ppi、rhi、三维散点图绘制和交互可视化。 主要功能: 1.单个体扫数据反射率因子各层仰角PPI图像; 2.单个数据反射率因子个方位角RHI图像; 3.某站点一段时间内连续数据0°仰角PPI图像连续显示; 4.单个体扫数据三位散点图交互可视化; 5.生成标准网格化数据。
3. Action Rec
.1. Vid2Doppler
Ahuja, Karan, et al. “Vid2Doppler: Synthesizing Doppler Radar Data from Videos for Training Privacy-Preserving Activity Recognition.” Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems. 2021. [pdf] [video]
.2. RadHAR
Singh, Akash Deep, et al. “Radhar: Human activity recognition from point clouds generated through a millimeter-wave radar.” Proceedings of the 3rd ACM Workshop on Millimeter-wave Networks and Sensing Systems. 2019. [pdf] [code]
.3. uDoppler-Classification
Classification of Human Movement using mmWave FMCW Radar Micro-Doppler Signature.
.4. mmpose
.5.mPose3D
.6. human_motion_recognition
implementation of the paper ‘Real-Time Human Motion Behavior Detection via CNN Using mmWave Radar’.
.7. mmWave-user-recognition
Janakaraj, Prabhu, et al. “STAR: simultaneous tracking and recognition through millimeter waves and deep learning.” 2019 12th IFIP Wireless and Mobile Networking Conference (WMNC). IEEE, 2019. [pdf]
4. Depth
.1. radar_depth
Lin, Juan-Ting, Dengxin Dai, and Luc Van Gool. “Depth estimation from monocular images and sparse radar data.” arXiv preprint arXiv:2010.00058 (2020). [pdf] [code]
.2. Simplified-2D-mmWave-Imaging
5. Vibration
.1. vital signs
生命体征就是用来判断病人的病情轻重和危急程度的指征。主要有
心率、脉搏、血压、呼吸、疼痛、血氧、瞳孔和角膜反射的改变
等等。正常人在安静状态下,脉搏为60—100次/分(一般为70—80次/分)。当心功能不全、休克、高热、严重的贫血和疼痛、甲状腺危象、心肌炎,以及阿托品等药物中毒时,心率和脉搏显著加快。当颅内压增高、完全性房室传导阻滞时,脉搏减慢。在一般情况下心率与脉搏是一致的,但在心房颤动、频发性早搏等心律失常时,脉搏会少于心率,称为短绌脉。
.1. TI 德州仪器
.2. 加特兰 [code]
.3. IMEC 微电子研究中心
Experiments have demonstrated the sensor’s ability for
multi-target detection
,heartbeat detection at 5 meter
and accurate tracking of a pedestrian’s position and velocity.
.4. Vayyar
Future Features Holisitic health and safety monitoring.
.5. mmVital-Signs
.6. WaveFace
- Face Verification
.2. Health Monitor
.1. Vayyar Home
Vayyar’s intelligent sensors
monitor location, posture as well as vital signs
, enabling behavioral monitoring such astime spent at rest
,in and out of bed
,nocturnal roaming
, andrestroom visits.
Trends are detected, allowing for pre-emptive predictions of health conditions such as UTI, dementia, and disorders like sleep apnea and psychological ailments including loneliness.
- Real-time fall detection
- Rich activity data collection
- Robust sensing in all conditions
- Maintains privacy
- Unobtrusive installation
.2. patient_monitoring
.3. mmWave_Gait
Meng, Zhen, et al. “Gait recognition for co-existing multiple people using millimeter wave sensing.” Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 34. No. 01. 2020. [pdf] [code]
.3. Fall Detection
.1. mmfall
6. Position&Tracking
.1. 房屋面积测量
根据AiP毫米波雷达具有距离高分辨率这一特性,结合波达角估计算法、恒虚警率检测及相应的边界搜索方法,最终得到了一种实用鲁棒的房间面积测量方法: 例如,通过对房间
面积的测量
,空调可以更好地控制制冷量,提高运行效率;电视、音响等家电可根据房屋面积以及人员有无调节音量大小
.2. Tracking
加特兰毫米波雷达室内人员检测与跟踪应用是基于60GHz/77GHz毫米波雷达芯片研发,采用FMCW、MIMO等技术,具有距离精度高、速度精度高、角度分辨率高及虚警率低等优点,可以实现室内情况下对
人员的准确检测
、精确定位和稳定跟踪
,并有效分类人与非人物体
,统计室内人员个数
,稳定输出人员的距离、速度和角度
等信息。
.3. mmWave-localization-learning
Deep Learning Architectures for Accurate Millimeter Wave Positioning in 5G
- An ML-based algorithm that enables energy efficient accurate positioning from mmWave transmissions, with and without tracking. [pdf]
7. Dataset
.1. DeepMIMO-codes
Ahmed Alkhateeb, “DeepMIMO: A Generic Deep Learning Dataset for Millimeter Wave and Massive MIMO Applications,” in Proc. of Information Theory and Applications Workshop (ITA), San Diego, CA, Feb. 2019.
a MATLAB code package of the DeepMIMO dataset generated using Remcom Wireless InSite software. The DeepMIMO dataset is a publicly available parameterized dataset published for deep learning applications in mmWave and massive MIMO systems.
Resource
- 降雨预测:https://github.com/yaoyichen/CIKM-Cup-2017
- mmwave 公司: https://www.calterah.com/news/
- Vayyar Imaging: a global leader in 4D radar imaging technology,
- Vayyar 公司: https://blog.vayyar.com/
- 学习资源: https://github.com/RadarCODE/awesome-sar 待学习