The Research And Application Of The Region Based Electric Bicycle Counting Algorithms

With the boost of modernization construction and development of computer hardware and software, the Intelligent Transportation System has been applied popularity. The road of city becomes more orderly and normal by applying the intelligent video processing technology. As an important part of city’s transportations, electric bicycle has been used by many people because it is convenient and cheap. While many problems have been caused by the irregular running of electric bicycle, so it is important to research and find an effective method which can detect and count the numbers of electric bicycle as a part of the Intelligent Transportation System.As one of the basic technology of video process, moving object detecting is the precondition of vehicle counting. The method of backgrounds subtraction can be divided into two classes:pixel based and region based, pixel based method deals with single pixel, so it is easy to appear noises. While, the region based algorithms can improve the detecting results by obtaining more information from the adjacent pixels. Two kind of region based moving object detection method are proposed in this paper to detect electric bicycle in real time. Base on this two moving object detection algorithms, an electric bicycle counting method with whole lane intersection correcting is discussed.The research content of this paper is mainly as follow.(1) Firstly, the Hu moment algorithm is analyzed, then the Gaussian Mixture Model of Hu moment indicator in moving object detection is presented and the performances of detecting electric bicycle are evaluated. Meanwhile, the ssim algorithm has been analyzed, then the Gaussian Mixture Mode of ssim is presented and the performances of detecting electric bicycle are evaluated as well. The experiment results show that the Hu moment indicator and ssim both can detect electric bicycle well, the Gaussian Mixture Model of Hu moment indicator and the Gaussian Mixture Mode of ssim both can detect electric bicycle as well. What’s more, the Gaussian Mixture Model of Hu moment indicator and Gaussian Mixture Mode of ssim have improved the adaptive capacity as for the varying scene.(2) Secondly, an electric bicycle counting method with whole lane intersection correcting is discussed, then it is applied for counting electric bicycle combined with the Gaussian Mixture Model of Hu moment indicator and the Gaussian Mixture Mode of ssim. The experiment results show that the detecting accurate rate and counting accurate rate of the Gaussian Mixture Model of Hu moment indicator and the Gaussian Mixture Mode of ssim are both higher than 95%.(3) Finally, an electric bicycle counting application software system is realized by applying C# language. the accelerating technique of CUDA and CPU accelerating technique have been used for optimizing the algorithm presented in the paper, making the counting accuracy and real time both meet the requirement of application.