Research On Small-footprint ALS Full-waveform Data Processing Technology

Airborne Laser Scanning(hereinafter, ALS), which is an active remote sensing technique, comprised of a number of sub-systems, with the characteristic of the real time and quick, high precision, good penetration, etc. The small footprint full-waveform ALS(footprint diameter less than 1m) can obtain more fine detail and higher accuracy of ground objects, compared with the traditional large footprint full-waveform system. It has great application value in many fields, like Agriculture, Forestry, Power, Military, etc. The small footprint ALS full-waveform data, resulted of many targets’ echo superposition, due to the characteristic of finer collection, and coupled with the effect of different kinds of complex noise at the same time. So the demands of processing method for small footprint full-waveform ALS waveform data is relatively high. It usually includes steps as preprocessing, waveform decomposition, constituents information(3D point cloud, intensity, wave width, etc) calculation and so on. The difficult point is preprocessing and waveform decomposition among them.The current widely used mehod, to process small footprint ALS full-waveform data, Gauss decomposition method and Deconvolution method, have the shortcomings that it can’t balance the denoising effectiveness and waveform feature preserving very well and the effect of waveform decomposition is poor. In this paper, we select the Deconvolution method, with stronger waveform decomposition capacity, through the research on small footprint ALS full-waveform data processing technology. The Deconvolution method, based on RL algorithm(hereinafter RL deconvolution method), has the better performance at stable, estimation for low SNR and edge detection. But it has the defect like convergence slowly, noise amplification and so on. And it will appear illusive wave crests, when we use the RL method to decompose the amplified wave width with too many iterations, thus the accuracy of waveform separation is influenced. In this paper, we make some improvements in the preprocessing and waveform decomposition, aimed at the defects.(1) The wavelet threshold denoising algorithm is used, in preprocessing. And according to the fact of small-footprint ALS full-waveform data processing, we realized the optimal selection of wavelet parameters, to improve the ability of denoising, so the influence of noise on RL deconvolution method is reduced, and more waveform characteristics is retained.(2) In order to reduce the influence of wave width on the accuracy of RL algorithm, we use the method, setting the cut-off rate of change on convergence curves, to realize the control of iteration on RL deconvolution effectively.(3) In order to enhance the convergence speed of RL algorithm, we introduce the accelerate RL algorithm based on two order vector extrapolation acceleration algorithm, after the construction quality improving of point spread function. The speed of small-footprint ALS full-waveform data decomposing by RL algorithm is risen nearly 6 times.At the last, the improvement effect of RL deconvolution method is verified with two sides, the comparative analysis of the effect of small footprint ALS full-waveform data processing and the application of target extraction.