Imaging light detection and ranging (LiDAR) utilizes the range values of target point clouds (or pixels) to estimate the three-dimensional (3D) shape of a target. However, ambiguity arises in determining which point clouds (or pixels) belong to the target when it is in cluttered or partially occluded environments, leading to uncertainties in reconstructing the 3D shape. Polarimetry, which examines the interaction between polarized light and materials, is a promising method for improving the determination of target pixels (or point clouds). The Mueller matrix, a comprehensive description of polarization properties, has been demonstrated as a powerful tool for characterizing and distinguishing targets with varying polarization characteristics. However, these demonstrations of Mueller matrix imaging have primarily focused on biomedical applications and have employed charge-coupled device time-sharing imaging. Additionally, the polarization diversity imaging technique has only been reported in synthetic aperture radar systems, based solely on the polarization imaging of target scattering matrices. We propose a multichannel polarization diversity imaging LiDAR that integrates Risley-prism multibeam scanning, frequency-modulated continuous wave (FMCW) coherent detection, and Mueller matrix polarimetry techniques for target classification. The proposed system eliminates ambiguity in determining which pixels (or point clouds) correspond to the target, a challenge faced by conventional imaging LiDAR systems.

To reduce artifacts caused by time-sharing polarization imaging, spatial parallelism and beam synchronization techniques are employed to simultaneously capture multipolarization point clouds using Risley-prism multibeam scanning. To achieve high spatial resolution in point clouds, the system incorporates the FMCW coherent detection, which offers high sensitivity, superior resolution, and a large dynamic range while simultaneously measuring distance and velocity. To address the issue of insufficient dimensionality for target recognition in unstructured environments with conventional imaging LiDAR, the Mueller matrix is introduced into LiDAR systems for target identification and classification. Different objects are distinguished and classified by leveraging variations in their polarization parameters.

A long-range 3D polarimetric imaging experiment was conducted using the proposed LiDAR architecture under outdoor conditions. The multidimensional information acquisition capability of the multichannel polarization diversity imaging LiDAR was exploited, and various filtering methods were applied to process the raw point clouds with a large field-of-view angle of ±30° (Fig. 3). To further demonstrate the LiDAR’s polarization imaging capability, small-scale buildings (A6 and A7) were selected as targets for recognition and classification (Fig. 4). In the cross-polarized point clouds [Fig. 4(b2)], buildings A6 and A7 were roughly divided into two regions: one with polarized reflectance intensities of 53?100 (colored in green) and the other with polarized reflectance intensities of 210?320 (colored in blue-violet). This indicates that the facades of buildings A6 and A7 are composed of two materials with distinct polarization properties. For more precise recognition and classification of building materials and structural details, Mueller matrix imaging was employed. The facades of buildings A6 and A7 exhibit a clear demarcation line (Fig. 5). Additionally, there is no ambiguity in determining which point clouds (or pixels) corresponded to the target. Different polarization parameters were utilized to distinguish and classify various targets, effectively enhancing the accuracy of target identification (Figs. 6 and 7).

We demonstrated both theoretically and experimentally a multichannel polarization diversity imaging LiDAR based on Risley-prism multibeam scanning, FMCW coherent detection, and Mueller matrix polarimetry techniques for building target recognition and classification. The multichannel polarization diversity and Mueller matrix characterization techniques were applied for the first time in imaging LiDAR. The proposed polarization diversity imaging LiDAR enables simultaneous multichannel distance, velocity, and intensity measurements, significantly reducing artifacts caused by time-shared polarization imaging. After noise filtering, building materials were initially distinguished from polarized reflection intensity point clouds for long-range targets. To achieve more accurate recognition and classification of building materials and structural details, the Mueller matrix and its associated parameters were utilized to characterize targets with different polarization properties. The proposed system resolved ambiguity in determining which point clouds (or pixels) belong to the target. Our architecture fully leverages the potential of 3D polarization imaging LiDAR for long-range target detection and classification in unstructured building environments, providing a promising approach for improving the accuracy of 3D perception in urban building target detection and classification.