基于无线感知和集成学习的人流量监测方法

Pedestrian flow monitoring method based on wireless sensing and ensemble learning

针对传统基于RSS信号进行人流量监测统计精度不高的问题,提出一种基于集成机器学习模型的人流量监测方法。采用部署在“门”型框架上的多个物联网节点组建的无线感知网络采集RSS信号,对其进行预处理提取特征值,基于“学习型”结合策略,构建由SVM、GBDT、XGBoost这3个初级学习器和逻辑回归算法作为次级学习器的集成学习算法模型来识别瞬时人数。实验结果表明,该方法能够大幅提升识别精度,达到了较高精度的人流量监测。

Aiming at the problem that the statistical accuracy of traditional traffic monitoring based on RSS signal is not high,a traffic monitoring method based on integrated machine learning model was proposed.A wireless sensing network composed of multiple internet of things nodes deployed on the“door”framework was used to collect received signal strength(RSS)signals.The feature values were extracted from the collected RSS.The stacking ensemble method was utilized,a two layers machine learning algorithm was designed to identify the instantaneous number of people.The first layer of the algorithm was constructed with three primary learners(SVM,GBDT and XGBoost),the second layer was a logistic regression algorithm.Experimental results show that the proposed method can greatly improve the recognition accuracy and achieve high-precision pedestrian flow monitoring.