MODIFIED SVD-PCA TECHNIQUE FOR TRAFFIC INCIDENT DETECTION

Abstract

In Real time traffic incident detection is acute for increasing safety and mobility on freeways. There has been incident detection approaches built on traffic behavior or mathematical models projected for this task. Though, earlier incident detection methods are partial in unique recurrent and non-recurrent congestions. The difficulty of current methods makes them insufficient to handle the real time task. In this paper, a novel approach for detecting incidents is proposed. The algorithm properties of singular value decomposition (SVD) and principle component analysis (PCA) matrix whose elements are local energies of coefficients at different scales. It uses the diagonal, left and right singular matrices obtained in SVD to determine the number of scales of self-similarity for road density, no. of vehicle and speed limit of location and scales of anomaly in data, respectively. Our simulation work on UK based authentic data sets validates that the technique achieves better detection rate that can be able to find if results nearly to zero for true and false accident rate for our hybrid approach NB based SVD technique better than the existing approach for self-similar data used C#.net based application.