Research Of Smart Parking System With Parking Spaces Prediction

Recently,with the rapid growth of motor vehicle ownership,the problem of parking has become obviously prominent.There are many reasons for the difficulty of parking,such as the occlusion of parking spaces information,the low efficiency of manually paying parking fees and the low utilization rate of parking spaces.In order to alleviate the parking pressure,it is a general trend to utilize advanced Internet technology and intelligent equipment to upgrade the traditional parking system.Traditional parking systems can not provide real-time parking information,do not support to pay parking fees online and can not predict parking spaces.With respect to these problems,based on the traditional offline charging software with license plate recognition system,we design and develop a smart parking system with parking space prediction on Android platform.By our smart parking system,users can inquire the real-time parking information,pay fee online and book parking spaces.During the developing process,based on the previous survey of users and parking operators,we firstly analyze the system requirements in detail and discuss the related framework and technique needed in the system.The communication protocol is HTTP and the format of data interaction is JSON.And then,the whole system is divided into several modules such as parking,parking assistant,personal center,etc.Furthermore,we design data tables according to the functional modules,and implement the system function according to the mode of Model-View-Controller.Meanwhile,in order to improve the utilization rate of parking spaces and the prediction accuracy of parking spaces,we design the methods to predict the unoccupied parking spaces based on the grey model and the neural network,respectively.Moreover,we optimize the neural network by dynamically adjusting the learning rate according to the change of deviation.To verify the effectiveness of these methods,the parking space prediction experiment is executed for one parking lot in Chengdu and the simulationresults are discussed.Finally,the optimized neural network by dynamically adjusting the learning rate is adopted as the prediction method in our system.At last,we carry out the functional test,non-functional test and user satisfaction survey.The results demonstrate that the system has the characteristic of good real-time performance,robustness,compatibility and high user satisfaction.The adoption of this system results in the more transparent information of parking space,higher parking space utilization rate,more flexible ways of paying and less time of finding parking space.Furthermore,the situation of traffic jam is indirectly improved and the environmental pollution is reduced to some degree.