Objective With the development of light detection and ranging (LiDAR) imaging technology, LiDAR imaging from monochromatic detection to multihyperspectral detection has developed into a new trend. Monochrome laser radar point clouds, different from traditional camera or multispectral image data fusion for visual color point cloud, the continuous spectrum of a laser as the light source of multispectral laser radar system can be synchronous detection more spectral channels of the echo signal, which can be directly obtained, have spatial and spectral information integration of point cloud data. This new type of point cloud data, which contains space spectrum information, lays a foundation for realizing three-dimensional point cloud color visualization imaging. Besides, it suggests higher requirements for quality control of the point cloud. This study uses intensity correction, color reconstruction, and color optimization methods to process the point cloud data of the multispectral LiDAR to obtain high-quality visual color point clouds. We believe that our method and research results can be useful for studying the direct acquisition of visual color point clouds by LiDAR.
Methods In this study, the point cloud data of our self-developed multispectral LiDAR is considered the research objective. First, an intensity correction model based on the LiDAR equation is proposed. By setting model parameters suitable for multispectral LiDAR, the influence of the angle and distance effect on point cloud intensity is corrected to lay a foundation for visual color point cloud reconstruction. Then, according to the theory of color reconstruction and considering the characteristics of the laser light source spectral power, detection wavelength, and the photoelectric response of the multispectral LiDAR system, a method of using the corrected intensity to reconstruct the color of the point cloud is proposed to realize the color reconstruction of the multispectral LiDAR point cloud. The color of the reconstructed point cloud is optimized using the polynomial regression algorithm to achieve the acquisition of a high-quality visual color point cloud. The color difference analysis of the point cloud before and after optimization is conducted to verify the proposed algorithm feasibility.
Results and Discussions The intensity correction model of the multispectral LiDAR is established to effectively correct the distance and angle effects on the intensity. The regression accuracy of each channel correction model is above 0.98 (Fig.2). After the original point cloud correction of the target in the scene, the point cloud strength is significantly enhanced, and the details are highlighted overall (Fig.3). Through the color reconstruction method, the point cloud intensity information of the corrected target is used to obtain the color point cloud that approximated the real scene (Figs.4 (a)). The polynomial regression algorithm is used to obtain the optimal mapping matrix from a series of training colors using parameterized functions. Besides, the matrix is used to optimize the color information of the point cloud, further improving the quality of the color point cloud (Figs.4 (b)). The color difference analysis shows that the color difference between the optimized point cloud color and the real scene reduces to the error range acceptable to the human eye (the color difference is less than 10) (Fig.5).
Conclusions Compared with the fusion of single wavelength LiDAR and passive image data to obtain visual color point clouds, multispectral LiDAR has a new approach for obtaining color point clouds. The multispectral LiDAR with multiwavelength and multichannel detection can directly obtain the point cloud data with integrated space-spectral information of the target. Aiming at the point cloud data of the multispectral LiDAR, this study first proposes the intensity correction model of the point cloud data of the multispectral LiDAR, and the feasibility of the model is verified by the measured point cloud data. which lay a foundation for color point cloud reconstruction. Besides, a color reconstruction and optimization method for a point cloud is proposed. The experimental results show that the color point cloud with a certain visual effect is obtained using the color reconstruction and optimization method for the point cloud. The optimization algorithm can significantly improve the visual effect of the multispectral laser point cloud. In the future study, the accuracy of color reconstruction of high point clouds will be further improved, making the reconstruction method more applicable to the reconstruction of complex scenes.
