A waveform decomposition technique based on wavelet function and differential cuckoo search algorithm

Waveform decomposition is widely used for the separation of echoes from full-waveform LiDAR (FWL) signal, and some previous studies employed Gaussian function for laser pulse modeling and waveform decomposition. However, it was difficult to guarantee the waveform parameters on the neighbor of optimal solution, because of the limited amplitude range. In addition, waveform parameters were usually set by the amplitude and location of inflection points, which may enlarge the difference between decomposed and original waveforms. Hence, a novel waveform decomposition technique based on wavelet function and differential cuckoo search algorithm is proposed, where wavelet function has a high-order vanishing moment, cuckoo search algorithm has a strong optimization ability, and differential operator avoids trapping into the local optima. The proposed technique is tested on airborne FWL point cloud and compared with other corresponding approaches, experimental results demonstrate that the decomposed waveforms are obtained with a reasonable convergence rate and feature characterization, as the rRMSE is lower than 7% for all of waveforms, the whole process of waveform decomposition only takes 0.3s, and waveform parameters are used as the features to recognize different objects from point cloud.