Research On Multi-domain Heterogeneous Traffic Data Fusion Model And Application

With the rapid growth of the world's population and the rapid development of urbanization,the issue of crowd safety in large public places has caused widespread social concern.On the one hand,due to the large spatial scope of large public places,the method of estimating the number of people based on single-view video no longer meets the counting needs of wide-area places.On the other hand,the rapid development of the Internet has spawned more intelligent sensors,giving data multi-domains,massive,and heterogeneous characteristics,which brings great challenges to data mining.In this context,this paper proposes a multi-domain heterogeneous traffic data fusion model,which aims to coordinate the accuracy of crowd state detection with the adequacy of multi-domain data utilization.The specific research content includes the following aspects:(1)Low-altitude local crowd counting method based on crowd gathering pattern.The multi-column convolutional neural network(MCNN)was used to estimate the number of people in the low-altitude perspective.Experimental results show that the method has certain accuracy.Combined with the personal space theory of psychology,the two maintain a high degree of consistency in the expression form of the crowd density distribution;(2)High-altitude global density estimation method based on Gaussian mixture model(GMM).A density grading algorithm based on a Gaussian mixture model is designed,and the crowd density in the high-altitude surveillance video is graded and displayed by the distance between key moving points.Experiments show that the method has a good performance in detecting crowd density;(3)Spatiotemporal evolution of crowd situation and early warning visualization system.Based on the Endsley situational awareness model structure,this paper designs a crowd situation visualization system based on Baidu Map API and applies it to actual scenarios,achieved good results;(4)Crowd counting method based on multi-domain distributed data fusion.Considering the limitation of the single-view crowd counting in wide-area venues,a distributed data fusion method is designed in this paper.Through the complementary fusion of local information and global information,the overall number of people can be estimated more accurately,and the prediction of the local crowd at the next moment is compared with the actual situation,and the result maintains a high accuracy.