Driving Behavior Analysis Based On Cloud Computing

With the sharp increase of car ownership, people pay more attention to the performance of the vehicles, traffic travel patency, driving safety, comfort and economy. Driving behavior analysis has become a common research for the vehicle and traffic fields. In recent years, the rapid development of cloud computing has provided a strong technical support for the analysis of driving behavior, which solves the problem that the traditional driving behavior analysis method cannot deal with the massive vehicle driving data. The analysis of driving behavior based on cloud computing are of great significance to the research of driving assistant system.The main work of this thesis includes the following several points:(1) The features which are closed related to the fuel consumption and safety has been extracted according to the Pearson correlation coefficient. The K-means, KFCM and the spectral clustering algorithm based on graph theory are illustrated in detail, and are used to analysis driving behavior. This approach obtains three categories from point of fuel according to the clustering results, including high fuel consumption driving behavior, moderate fuel consumption driving behavior, and low fuel consumption driving behavior. And obtains three categories from point of safety, including aggressive driving behavior, moderate driving behavior, cautious driving behavior.(2)Cloud computing platform has been built based on the virtualization technology of OpenStack and Spark parallel processing technology. Parallel spectral clustering algorithm has been realized based on cloud computing platform for handling the big data to analysis the driving behavior.(3) The core principle of BP neural network algorithm has been researched Meanwhile, the training data based on the clustering results is constructed Then, the online driving behavior model through BP neural network is established to perform online evaluation.(4) Cloud computing platform has been realized by configuring the software and hardware environment. Driving behavior was analyzed based on parallel spectral clustering algorithm. And then we finished the verification for driving behavior evaluation model based on BP neural network.