Realization Of Intelligent Transportation System Based On Hadoop Open Source Framework

With the continuous development and application of Internet of Things technology, intelligent city, intelligent transportation and other new concepts designed to improve the municipal construction, people's lifestyles have penetrated into all aspects of urban life, but most of the existing intelligent traffic system for the traffic control department, The use of government agencies, the lack of open data interface for the public,can provide a single form of service lack of practicality, while the traditional data storage methods have many shortcomings, there are low data maintenance efficiency, high hardware costs and other issues, which are certain To the extent that the feedback of user information is slow and can not reach the initial goal of intelligent transportation. At the same time,the data mining of mass heterogeneous data is mainly based on the improvement of data mining algorithm in the traditional stand-alone environment. Due to the performance factors such as computer memory, it can not meet the processing requirements of the constantly increasing perceptual data. Based on the shortcomings of the current intelligent transportation system, this paper makes the work in three aspects: system platform performance, data sharing mode, historical data mining and so on.(1) First, in order to efficiently collect and store heterogeneous data,this paper achieves the goal of easy management and resource saving. This paper realizes the data receiving module which combines asynchronous non-blocking IO and message queue. Based on the idea of reading and writing separation, Multiplexers are responsible for managing connections,reading data, and writing data. Message queues are used to eliminate the unbalanced processing speed between the upper and lower messages, so that many sensing nodes can interact with the back-end server smoothly and ensure the system Timeliness and robustness. Due to the heterogeneity of the data, for different types of data characteristics and the use of different design relational database mysql storage of historical data, non-relational database MongoDB real-time data storage to improve data read and write efficiency, but also to a large extent Saving hardware costs. In addition,because the main way to share data is web service, in order to meet the requirements of web response and high concurrency, this paper proposes a distributed two-level cache system, which is shared by Ehcache distributed cache and Redis centralized cache. , Shortening the length of the user response processing chain, and improving the response speed.(2) Secondly, in order to make full use of data, this paper designed the following data sharing interface: 1. For the environment-aware data using real-time table retrieval and historical data charts show two ways. 2. For the LBS geographic location information, real-time tracking of vehicle location and human flow heat map are adopted. 3. Vehicle real-time video surveillance and picture display. Through these basic interfaces to complete the external data sharing, improve system service levels.(3) Finally, in order to satisfy the user's travel demand, the paper realizes the K-Nearest Neighbor short-term traffic flow forecasting module based on Hadoop platform by mining the deeper correlation of the data,and designs reasonable state vector, distance vector and prediction function to improve the prediction Accuracy, the MapReduce parallel computing framework is used to solve the problem that the K-nearest neighbor algorithm is inefficient in the large data condition, and the timeliness of the short-term traffic flow forecasting is improved, and the whole efficiency and efficiency of the scheme are validated by experiments.