Method for Removing Multi-Frequency Vibration Information From 3D Laser-Scanned Pavement Point Clouds

Large-area 3D laser scanning is susceptible to interference from multi-frequency vibration noise of an acquisition vehicle, resulting in low accuracy of the acquired 3D pavement morphology. Traditional filtering and image processing techniques cannot perform component analysis and complex processes. Thus, a variational modal decomposition (VMD) algorithm based on the modified Harris Hawk optimization (AMHHO) algorithm is proposed to analyze the pavement components and achieve accurate stripping of multi-frequency vibration information. The pavement point cloud data acquired by the vehicle-mounted 3D laser camera is downscaled to obtain the pavement longitudinal profile signal. This signal is then decomposed by the proposed AMHHO-VMD algorithm to obtain intrinsic mode functions, which are then Fourier-transformed and combined with the vibration state of the acquisition unit to determine the multi-frequency vibration information. Finally, the accurate 3D morphology of the pavement is obtained after reconstruction of the filtered effective components. Experimental results show that compared to the empirical mode decomposition (EMD) algorithm and wavelet packet decomposition algorithm, the proposed AMHHO-VMD algorithm can strip the multi-frequency vibration components from the original pavement point cloud and obtain an accurate 3D morphology of the pavement.