Traffic surveying systems using pneumatic road sensors are currently widely used in Australia for counting and classifying vehicles. However, these intrusive sensors disrupt traffic and expose technicians to significant road dangers. We propose a non-intrusive automated vehicle classification system for the AUSTROADS classification standard based upon laser sensor technology. The proposed system is capable of classifying vehicles in multi-lane, high-speed environments. Conventional Fourier-based denoising techniques are, however, unable to meet the design challenge due to a considerable amount of noise presented in measurement data that is induced by both the laser sensing device and high volume traffic in carriageways. This paper proposes an advanced wavelet-based denoising technique to greatly enhance the noise reduction performance of the proposed automated vehicle classification system.