The objective of this research is to develop a novel non-contact damage detection system for cracks located in railhead since such cracks are responsible for a significant portion of railway accidents. The system performs non-contact measurements by using two laser Doppler vibrometers (i.e., LDV’s) placed on a moving rail car which records the vibrations of waves propagating in the railhead induced by moving rail cars. The recoded signals are processed in the frequency range between 30 kHz and 100 kHz. This is because, in this range, waves localize at different sections of rail such as railhead, web and foot. Therefore, LDV measurements record the vibrations of waves propagating in the railhead. To investigate the interaction of the waves with the cracks, Finite Element Method (i.e., FEM) simulations were adopted. The main challenge is to minimize the speckle noise which is inevitable in moving LDV measurements. Speckle noise causes impulses in the time domain and broadband noise in the frequency spectrum. To overcome the former problem, the methods to detect impulsive noise (i.e., IN) corrupted segments and the methods to estimate the lost signal segments due to IN are selected from the literature and further modified to enhance their performance. Combination of the best performing methods led to a robust IN filter. To eliminate the broadband noise component, approaches such as input-output identification, generation of the green function using two measurements, and wavelet-based noise reduction methods will be implemented. The goal to perform damage detection at operational speeds of rail cars.