Research On Vehicle Postures Recognition Method Based On Multi-source Sensor Information Fusion At The Overpass

Multilayer road networks,such as overpasses and elevated roads,have been increasingly used to solve traffic congestion problems in large cities.At the same time,people pay more and more attention to the role of three-dimensional navigation technology in the process of urbanization and informatization.How to get the depth information of vehicles to distinguish the position of vehicles in the three-dimensional traffic network and to make accurate navigation routes has not been well solved.When vehicles use the navigation system in multi-level road networks,it is very important to identify the upper and lower positions of vehicles on different levels of roads,especially for overlapping complex overpass roads.Although they can be represented and visualized in the existing navigation system,at present,because the existing vehicle positioning system uses ordinary consumer GPS,and Geographic Information System for Transportation Database is mainly based on two-dimensional navigation.In multi-layer road network,due to the height error of handle GPS,it is difficult to determine the vertical position of vehicles on different roads.This dissertation presents a method of using smart phone sensor information fusion to distinguish the vehicles postures in the overpass section,based on the deficiencies of existing vehicle navigation with consumer GPS in overpass road network.The typical overpass has complicated shape,variable layers,and the postures of vehicles are only categorized as nine species,such as up-down,left-right and straight-up,at the changeable overpass.Its basic idea is to judge if the vehicle has be entering the overpass section by consumer GPS data with existing positioning accuracy,and collect the information of pressure,direction and acceleration by smart phone with sensor collector during the vehicle driving.The initial sensor information with noise is denoised by the WT(wavelet transform)de-noising technology,and the data dependence analysis,feature extraction,using mature Machine Learning SVM algorithm to build a prediction model which intelligently determines the vehicle's up-down left-right direction.The number of each section and ramp of several typical overpasses in Shenzhen is marked,and the structural information of the overpasses is constructed.Then,the two-dimensional vehicle trajectory data can be matched with the position data of the overpasses network section,and the road position of the overpass can be judged by analyzing the changing trend of vehicle postures data.This dissertation mainly concerns on the key technologies related to the postures of vehicles in urban overpasses,to improve the accuracy of heading information of vehicle in complex overpasses sections.The experimental results show that the method has high accuracy about the postures recognition of the vehicle in the interchange section.To a certain extent it can serve solving the chaotic situation of vehicle positioning in the navigation of the urban overpass section.