Research On Detection Method Of Abnormal Pavement Variation By Fusing Trajectory And Mobile Phone Multi-sensors

Road is an important foundation of transportation and plays an important role in national construction and economic development.With the economic and social development and the increase of vehicle ownership year by year,the road surface burden is larger.In addition,due to the delay of road maintenance and other reasons,the road surface has a large number of abnormal characteristics,and has a greater impact on safe driving and vehicle life.Therefore,in order to quickly obtain the abnormal change information of the road surface,it has become an urgent need to be solved by the traffic department.Traditional road surface information collection methods have problems such as long data collection cycle and high cost.Therefore,this paper proposes a detection method for abnormal changes of urban road surface that integrates trajectory and mobile phone multi-sensors.By using the sampling data of sensors such as position,acceleration,attitude and magnetic field built into the mobile phone,the research is carried out from three aspects of sensor attitude solution,construction of BP neural network for detection of abnormal road surface changes,and detection of abnormal road surface changes:(1)the data in the attitude is important step before analyze the data,this article will analyse the representation method for the attitude Angle choose the representation of small amount of calculation,low error probability,and study the general model of the attitude algorithm algorithm,finally to the characteristics of the mobile sensor data,attitude algorithm based on kalman filter algorithm is put forward.(2)According to the characteristics of abnormal changes in the pavement of the object studied in this paper,the appropriate selection neural network model is selected to construct the neural network structure suitable for the object studied in this paper.In addition,the appropriate site and the appropriate expected value are selected as the training connection of the neural network,and the problem of abnormal surface object detection on the pavement is finally solved.(3)By comparing and analyzing the original collected data with the attitude correction data,the effect of attitude calculation and correction is verified.The detection results of threshold method and neural network are compared and analyzed to verify the effect of neural network in detecting abnormal changes of road surface.The method presented in this paper has low requirements on hardware equipment and high data collection efficiency,which reduces the cost of road surface information collection.The experimental results show that the method presented in this paper can detect the abnormal changes of road surface quickly and accurately,and the accuracy of the result is more than 85%,which has a broad application prospect.