In this study, wavelet transform was firstly applied to deal with a numerically simulated signal that was unable to obviously identify axle information. The analysis result showed that the wavelet transform was able to magnify the slope discontinuities so as to accurately identify the silhouette of passing vehicles. Subsequently, based on the field-tested FAD signals through which the vehicle configuration was difficult to be directly identified, the most appropriate transform scales and the best suitable wavelet function performing wavelet transform were selected from the minimum Shannon entropy and maximum correlation. The results demonstrated that the wavelet transform in pattern recognition effectively identified the vehicle configuration (including vehicle velocity, axle numbers, and axle spacing), especially for the unidentified FAD signals. Therefore, wavelet domain analysis can effectively improve the efficiency and accuracy for the vehicular axle identification in BWIM system, and it is beneficial for the successful application of BWIM system in controlling and monitoring overweight vehicles.