This research attempts to improve the accuracy of vehicle re-identification at successive loop detector stations through improving the distance measures in the pattern matching process. Vehicle inductance-signature data, collected by a California team of researchers, were further analysed at the University of Toronto. Several new distance measures were used to match the normalised waveforms that proved to be outperforming previous features. Other approaches such as horizontal shifting of the waveforms for warping-reduction and Back Propagation Neural Network were also investigated.
Abstract