Research On Highway Traffic State Estimation Method Based On Mobile Phone Data Features Denoising

Mobile phone signaling data provides excellent conditions for fine-grained and large-scale estimation of traffic states due to its advantages such as wide coverage,low collection cost,and large data volume.However,due to various complex factors,there is intensive noise in mobile phone data,which is more likely to affect the accuracy of estimation results of traffic state estimation models.Therefore,this thesis mainly focuses on the noise characteristics of mobile phone feature data,proposes an improved density peak clustering algorithm to filter the data noise,and establishes a traffic state estimation model based on mobile phone feature data,and uses denoised data to achieve high-precision,fine-grained,and wide coverage estimation of highway traffic state.This thesis firstly summarizes the application and research status of mobile phone signaling data in the field of transportation and traffic state estimation models.On this basis,the collection principle and preprocessing process of mobile phone data are summarized,and the mobile phone speed feature extraction method based on mobile phone signaling data is introduced.And the principle of data collection based on microwave radar is introduced.Aiming at the distribution characteristics of the original mobile phone speed feature data and the main source of data noise,this thesis proposes a denoising method based on the improved density peak clustering algorithm(DPCA).Considering that in the traditional density peak clustering algorithm,the cutoff distance parameter is generally set subjectively by humans,and the defect of artificially selecting the clustering center,which makes it difficult to achieve adaptive clustering and denoising for different data distribution.Combined with "Data field" and information entropy theory,this method optimizes the cutoff distance adaptive selection algorithm based on different data distribution characteristics.On this basis,based on the highway traffic state classification standard and the specific characteristics of the effective feature data category,this method optimizes the adaptive selection of clustering centers and the automatic selection of effective data categories,and realizes the effective filtering of data noise.On this basis,this thesis establishes a traffic state estimation model based on the mobile phone feature data.The model is based on the basic structure of LSTM,and uses the speed and quantity features of the mobile phone as input to estimate the traffic speed on the corresponding road section and driving direction.This thesis describes in detail the data flow of each layer in the model and the calculation process of each gate control structure.Finally,this thesis takes the Shanghai-Nanjing Expressway as a case study area,and uses relevant mobile phone feature data and radar data as the basis to verify the data denoising method and traffic state estimation model proposed in this thesis.The results show that the denoising method proposed in this thesis can effectively filter the data noise and reduce the influence of extreme noise data.And the traffic state estimation model based on filtered data training and application has a significant improvement in the estimation accuracy,and the output result of the model can reflect the actual traffic situation well.