Automated Vehicle Detection and Classification using Acoustic and Seismic Signals

Security threats to important infrastructure cause problems to not only those who live nearby but also in a much wider sense. It is therefore desirable to consider the use of automated systems capable of detection and identification of potential threats. This thesis describes an investigation into acoustic and seismic methods for achieving such a system specifically for commercial road vehicles. Accurate algorithms have been developed for recognition of moving vehicles using fusion of acoustic and seismic signals. It has been found that seismic signals are less susceptible to interfering signals, making them optimal for detection of vehicles. Their much narrower bandwidth also increases processing efficiency and speed. Thus, the algorithm developed utilises firstly only seismic signals to detect vehicle presence, and then employs both acoustic and seismic signals for classifying type of the vehicle. The detection algorithm is purely time domain and uses seismic Log Energy together with a modification of Time Domain Signal Coding. The best detection accuracy obtained was 97.71 % with Support Vector Machine and 99.02 % with Learning Vector Quantisation Neural Networks. The classification algorithm to distinguish between trucks and cars utilises three relatively simple time domain methods: Zero-Crossing Rate, Log Energy and Autocorrelation of seismic signals; combined with LPC coefficients collected from acoustic signals. Classification with either SVM or LVQ reached 93.30 % or 80.80 % respectively. This study therefore has demonstrated it is possible to detect an approaching vehicle and classify its type by using acoustic and seismic signal processing.

Name dc:type.qualificationname

Ph.D

Level dc:type.qualificationlevel

doctoral

Grantor dc:publisher.institution

University of York

Year dc:date.issued

2010

Author dc:creator

• Evans, Naoko

Advisor dc:contributor.advisor

• Chesmore, David

Identifier

uk.bl.ethos.557169

OAI identifier oai:identifier

oai:etheses.whiterose.ac.uk:1151