Research On Color Denoising And Highlight Removal Method Of Point Clouds For Multispectral LiDAR

As the main technical means of three-dimensional(3D)imaging,light detection and ranging(LiDAR)technology can obtain the spatial information of the detected target,thus forming high-precision point cloud data,which is widely used in smart city,unmanned driving,resource survey,cultural relic protection and other fields.Due to the lack of spectral information or color information acquisition ability,existing singlewavelength LiDAR is difficult to realize color 3D imaging,which limits the point cloud visualization ability.In order to solve this problem,many researches attempt to obtain spatial information and spectral information simultaneously from hardware or algorithm by combining active LiDAR and passive multispectral imaging.Although the technology has made some progress,there are still some problems such as illumination influence and point-surface matching.Compared with the traditional single-wavelength LiDAR,multispectral LiDAR(MSL)can realize the synchronous integrated acquisition of spatial information and spectral information,so as to construct a new kind of point cloud data--MSL color point cloud data.Compared with the monochromatic point clouds,the color point clouds not only have significant advantages in the application of target recognition and physical property discrimination,but also have greater application potential in point cloud visualization.In the process of color point cloud acquisition,it is inevitable to receive the influence of equipment,environment and human factors.Furthermore,it will interfere with the expression of geometric information and color information in point cloud data,resulting in point cloud noise.Due to the lack of color information in monochromatic point clouds,most of the existing researches on point cloud denoising focus on geometric scale noise removal,but ignore color scale noise removal.Therefore,in order to further improve the point cloud visualization effect of new MSL color point cloud data,this paper carries out research on color point cloud denoising method based on color information from two aspects of color noise and highlight noise removal.The main research contents of this paper are as follows:(1)The composition and principle of MSL system are introduced in detail.In order to improve the quality of the key signals of the system,the multi-channel echo signal control is designed from the aspects of power ripple suppression,signal amplification control and signal synchronization control.This ensures the effective acquisition of signals and lays a foundation for subsequent generation and processing of point cloud data.(2)The principle of color point cloud generation of MSL is described in detail from two aspects of multi-channel waveform information extraction and multi-spectral point cloud color reconstruction.Further,according to the causes of point cloud noise,color point cloud noise is analyzed based on color information,and the necessity of removing color noise and highlight noise for improving the visual effect of point cloud is derived.(3)The color point cloud denoising method based on color information is proposed.Firstly,in the aspect of color noise removal,in order to improve computing efficiency and ensure implementation effect,data dimension reduction based on feature retention was carried out to transform 3D point cloud into two-dimensional(2D)image.Then according to the statistical features of color noise,apply combined denoising method to remove two different color noise.Secondly,in the aspect of highlight noise removal,weak highlight and saturated highlight are recognized by threshold detection and visual saliency detection respectively,and the local highlight noise removal problem is transformed into color repair based on global variables.(4)The MSL scanning experiments are designed to verify the validation of the proposed algorithm.From the perspective of qualitative visualization effect and quantitative evaluation index,the algorithm in this paper can effectively and robustly remove color noise and highlight noise of color point clouds,thus improving the validity and accuracy of color restoration of point cloud and further improving the visualization effect of point clouds.