面向多数据流的车厢拥挤回归分析方法  

Regression analysis method of compartment congestion oriented to multiple data streams

作　　者：奚蓓灏 汪明明 陈庆奎[1] XI Beihao;WANG Mingming;CHEN Qingkui(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《计算机应用》2021年第S02期314-317,共4页journal of Computer Applications

基　　金：国家自然科学基金资助项目(61572325);上海市重点科技攻关项目(19DZ1208903);上海市工程中心建设项目(GCZX14014);上海智能家居大规模物联共性技术工程中心项目(GCZX14014)。

摘　　要：公交车拥挤度分析对维护公共交通安全起着重要的作用。针对在传统的目标检测方法中使用单个摄像头导致无法获取完整的车厢图片信息,以及在高密度场景下乘客与乘客之间的遮挡或者乘客被车厢内的座椅等物体遮挡的问题,提出了一种借助两个前后车厢的摄像头面向多数据流的车厢拥挤回归分析方法。首先,定义一个线性方程;其次,获取相对可见信息:公交车最大核载人数、根据人眼标记出的总人数、以及通过YOLOv3和ResNet50分别检测出车厢内人头数和拥挤率;然后,将包含已知信息的样本数据矩阵和期望值向量代入所定义的方程中,拟合出隐含信息:系数向量和偏置项,构建出一个多元一次线性回归方程,在高密度环境中狭窄和遮挡严重等情况下能够获得更为精确的车厢内总人数;最后,通过人数估计线性回归算法,获得最终的车厢内总人数。实验结果表明,所提方法能够预测出公交车上的人数,实时获得公交车上的人群流量,并且通过平均绝对误差(MAE)和均方误差(MSE)对数据进行误差分析后,验证了该方法能够正确地反映公交车拥挤度。The analysis of bus congestion plays an important role in maintaining public transportation safety. The traditional object detection methods use a single camera which can not achieve whole bus picture information,and passengers are usually blocked by other passengers or objects in the bus such as seats in high-density scenes. In order to solve the problems,a regression analysis method of compartment congestion oriented to multiple data streams with the front and rear compartment cameras was proposed. Firstly,a linear equation was defined. Secondly,the relative visible information was obtained:the maximum number of people checked on the bus,the total number passengers marked by the human eyes,and the number of passengers on the bus and the congestion rate respectively detected by YOLOv3 and ResNet50. Then,the sample data matrix and the expected value vector containing the known information were substituted into the defined equation to fit the hidden information:coefficient vector and bias term,and a multivariate linear regression equation was constructed to get more accurate total number of passengers on the bus in high density environment with narrow and severe occlusion.Finally,the total number of passengers on the bus can be obtained through the linear regression algorithm of number estimation. The experimental results show that,the proposed method can predict the number of people on the bus and obtain the crowd flow on the bus in real time. And through the Mean Absolute Error(MAE)and Mean Square Error(MSE)of the data error analysis,it is verified that the proposed method can correctly reflect the bus congestion.

关 键 词：公交车拥挤度 多数据流 回归模型 YOLOv3 ResNet50 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]