Millimeter-Level LiDAR Point Cloud Ranging via Optical Time Delay Asynchronous Sampling at Low Sampling Rates

Time-of-flight (ToF)-based light detection and ranging (LiDAR) systems are widely adopted due to their high system integration and broad adaptability. However, achieving millimeter-level ranging accuracy necessitates enhancing the time measurement precision of the time-to-digital converter (TDC) module to the picosecond (ps) even to sub-picosecond range, which significantly increases system development costs and power consumption, thereby limiting its adoption in low-cost applications. This study proposes an optical time delay asynchronous sampling method to improve ranging accuracy in ToF-based LiDAR, successfully achieving millimeter-level ranging precision under low-sampling-rate conditions. The approach effectively addresses the challenges of high system complexity, elevated hardware costs, and intensive data processing demands in high-precision ToF-based 3D point cloud mapping. A simulation model of optical time delay asynchronous sampling is first established to validate the feasibility of enhancing ranging accuracy through temporal signal shifts, low-rate sampling, and Gaussian function waveform fitting. Subsequently, an experimental validation system was established with a laser pulse width of 225 ps and a sampling rate of 6.25 GHz, which has the sampling interval intentionally set to approximate the pulse width of the laser echo signal. The impact of waveform fitting methods, optical time delay operations, sampling counts, and ranging distance on ranging accuracy is quantitatively analyzed and optimized. The optimal root mean square error (RMSE) and standard deviation of ranging measurements are achieved with four optical time delay operations, resulting in an RMSE of the point cloud of 0.1548 mm. Finally, 3D reconstruction experiments on millimeter-scale models demonstrate that the proposed method preserves height information with a 1 mm step difference, enabling precise recovery of surface details. This contrasts with traditional low-sampling approaches, which suffer from significant height data loss, offering a novel solution for simplifying system integration, reducing costs, and achieving high-precision 3D point cloud mapping under low sampling rates.