Research On Signal Detection For Airborne LiDAR Bathymetry

Airborne LiDAR bathymetry(ALB)has shown enormous potential in the field of shallow water and coastal mapping.However,it still has great difficulty in data processing,which restricts the further development of the technology.This thesis focuses on one of the key technologies—bathymetric signal detection—involving sea-land echo classification,waveform filtering,waveform deconvolution,peak detection,and waveform decomposition.The main work and innovation are as follows:1.The status of ALB technology is discussed.The composition of ALB system,its bathymetric theory,and the basic flow of data processing are presented.Waveform components and the radiative transfer model are analyzed.Appropriate improvements have been made to the processing flow for data characteristics.2.An automatic sea-land echo classification method based on multi-channel features is proposed.In this method,training samples can be automatically generated based on the nearinfrared echo features,and a new multi-channel feature extraction method is proposed,which overcome the defects of traditional methods that require manual sample collection and connect the unlabeled echoes with training samples.Four classifiers are applied to echo classification respectively.Experimental results prove that this method can achieve high classification accuracy without human intervention.3.To solve the three key issues in bathymetric signal detection—overlapping signals in shallow water echo,signal detection for weak bottom echo and effects of water column backscattering—the existing waveform processing methods are analyzed from four aspects,including waveform filtering,waveform deconvolution,peak detection and waveform decomposition,and the performance of related algorithms is assessed and compared,supporting the study of data processing technologies for ALB with important experimental results.4.To solve the problem that existing waveform processing methods have limited application,a depth-adaptive signal detection method for ALB is proposed.The water depth is fast estimated by the effective range of the waveform.In waveform preprocessing,the advantages of Richardson–Lucy deconvolution and average square difference function are effectively fused with respect to depth.In waveform decomposition,an exponential function with a second-order polynomial model is proposed to reduce the sensitivity to depth,and the trust region algorithm is introduced to solve the model parameters.Experiments show that this method not only improves the accuracy significantly,but also maintains better robustness for waveforms at different depths.