Line laser scanning with automatic exposure and CUDA acceleration

To address the challenges of fast and accurate line laser scanning in complex measurement scenarios, this paper proposes methods for automatic exposure and efficient centerline extraction of line laser scanning. The impact of exposure time on laser stripe images is analyzed, and an automatic exposure method is developed, which utilizes grayscale histogram statistics and iterative solutions to adjust the exposure time, thereby ensuring the grayscale distribution of the laser fringe images closely aligning with the theoretical Gaussian distribution. For centerline extraction, a method based on the Hessian matrix is proposed in conjunction with the non-maximum suppression strategy, effectively reducing data processing volume and enhancing the accuracy and efficiency of centerline extraction. Additionally, GPU acceleration techniques are utilized to further improve the execution speed of the algorithm. Measurements on objects made of different materials demonstrate the effectiveness and universality of the proposed methods. Compared with the Steger method, an overall measurement error of less than 0.01 mm is achieved while the centerline extraction time is reduced by approximately 50 %.