Laser Active Alignment Algorithm based on Spot Features and Curve Fitting

Abstract

This paper addresses the active alignment problem in manufacturing solid-state LiDAR systems and focuses on image processing and alignment algorithms within a visual laser auto-collimation platform. By extracting features from laser spot images, a non-linear discrete optimization is applied to determine the optimal position of the laser emitter, minimizing the laser beam divergence angle. To overcome the challenge of inaccurate size extraction of Fraunhofer diffraction patterns, Gaussian filtering, adaptive thresholding, and circle fitting based on distance transformation are employed to estimate the spot size. Active alignment algorithms are implemented using improved hill climbing, genetic, and curve fitting algorithms. Experimental comparisons demonstrate that the curve fitting-based active alignment algorithm achieves better efficiency and stability, with an average adjustment count of 29.35 and a variance of 97.50.