目前智能空调的研究主要基于视觉或红外传感器，基于单个4D毫米波雷达的相对较少。采用单个4D毫米波雷达传感器采集数据，根据不同目标反射点位置的不同提出了密度聚类(Density-Based Spatial Clustering of Application with Noise，DBSCAN)算法实现目标的聚类识别，并与K均值(K-means)聚类算法进行了效果对比。针对多目标跟踪问题，设计了一种基于联合概率数据关联算法(Joint Probabilistic Data Asso-ciation，JPDA)和卡尔曼滤波(Kalman Filter，KF)的目标跟踪算法，从而实现多目标的匹配和跟踪。将所研究的算法应用到4D毫米波雷达系统，并在室内采集了行人目标数据，分析对比实际场景和算法跟踪效果，误差大约在 8 cm内，准确率可达91。8%。结果表明:该算法可以较好地实现多目标跟踪，可用于智能空调中。
The Multi-Target Tracking Algorithm Based on 4D Millimeter-Wave Radar Point Clouds
Current research on intelligent air conditioners is mainly based on visual or infrared sensors,and research based on a single 4D millimeter-wave radar is relatively little.This paper uses a single 4D millimeter-wave radar sensor to collect data,proposes a density-based spatial clustering of application with noise(DBSCAN)algorithm to realize the clustering recognition of targets according to the positions of the reflection points of different targets,and compares its effect with the K-means clustering algorithm.For the problem of multi-target tracking,this paper de-signs a target tracking algorithm based on joint probabilistic data association(JPDA)and Kalman Filtering(KF)to realize multi-target matching and tracking.The algorithm studied in this paper is applied to the 4D millimeter-wave radar system,the pedestrian target data is collected indoors,and the actual scene and the tracking effect of the algorithm are analyzed and compared,with an error of about 8cm and an accuracy rate of 91.8%.The results show that the algorithm can achieve multi-target tracking well and can be used in intelligent air conditioners.
万方数据
维普
