Traffic Incident Detection Algorithm Based on Wavelet Support Vector Machine

Abstract

In order to improve the accuracy of modeling, a wavelet de-noising support vector machine (WD-SVM) is proposed based on the time-frequency property of wavelet transform and the superior classification function of support vector machine (SVM). First, the original data is de-noised using wavelet technology, and then adopted to construct SVM models which with Polynomial kernel, Radial basis kernel, and Sigmoid kernel respectively are used in real-time traffic incident detection on highways. So as to testify the validity and portability of the models, a simulation experiment is designed using I-880 data, and a combined index is also devised to evaluate the results comprehensively. The result shows that different kernels make different detecting results, and one of the most outstanding performance kernels is Polynomial. In addition, by comparison of the traditional SVM model and WD-SVM proposed in this paper, it finds that the WD-SVM shows a significant advantage on detection rate (DR), false alarm rate (FAR) and mean time to detection (MTTD).