Sparse Representation Denoising Framework for 3-D Building Reconstruction From Airborne LiDAR Data

作者：Cao, Zhimin; Gu, Yanfeng^*

来源：IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2016, 9(5): 1888-1900.

DOI：10.1109/JSTARS.2015.2448126

• Airborne LiDAR
• denoising
• sparse representation
• three-dimensional (3-D) building reconstruction

Random noise has a serious impact on the performance of three-dimensional (3-D) building reconstruction from airborne LiDAR data. In this paper, a sparse representation denoising framework for building roofs from airborne LiDAR is proposed. In the proposed framework, both the random noise and the local structural information are considered. At first, a systematic analysis for the random noise of the rasterized image of raw LiDAR data is presented in detail by taking the random noise of LiDAR data and its local structural information into consideration. With the proposed random noise model, the rasterized image can be adjusted into image with White Gaussian noise. Therefore, by adjusting the rasterized image with the proposed random noise model, sparse representation denoising framework designed for White Gaussian noise is employed in this paper. In order to realize the sparse representation-based denoising framework efficiently, multimanifolds structural dictionaries are learned from clean simulated data by employing K-SVD technique. Finally, three different implementations of the proposed denoising framework are proposed. Experimental results illustrate that the proposed denoising framework can efficiently restore the lost information caused by random noise of 3-D building roof data from airborne LiDAR with obvious improvement comparing with the classical K-SVD-based denoising method.

• 出版日期2016-5
• 单位中国石油大学（北京）; 哈尔滨工业大学