基于GWGCGM和DBSCAN的线结构光条纹中心提取方法

Centerline Extraction Method of Structured Light Stripe Based on GWGCGM and DBSCAN

为解决传统光条提取算法难以同时保证算法的抗干扰能力和运行效率的问题,提出一种高斯加权灰度重心法与密度聚类相结合的亚像素条纹中心提取方法,该方法对含有噪声点的图像能有效提取光条中心线。首先对图像用单通道提取、伽马校正、大津法做预处理,以提取图像中的有效光条纹信息;然后用高斯加权灰度重心法初步拟合光条纹中心线,再利用密度聚类算法进一步精确拟合光条纹中心线,最终得到中心线的亚像素坐标。仿真实验结果表明,在处理含有随机噪声点数小于1000的一组原始图像时,该算法均方根误差都在0.72 pixel左右,均方根误差最大波动为13.6%;与传统的Steger算法相比,该算法运行效率提高了4.8倍且精度仅降低了9.2%。

A sub-pixel stripe center extraction method leveraging the Gaussian-weighted grayscale center of gravity and density clustering is proposed.This approach addresses the limitations of traditional strip extraction algorithms,particularly their susceptibility to interference and the need for enhanced efficiency.The proposed method successfully extracts the center line of the light bar for images with noise points.The effective light-stripe information in the image is first extracted using the single channel extraction,gamma correction,and Otsu method.The light-stripe centerline is then initially fitted using the Gaussian-weighted grayscale center of gravity method.Finally,the light-stripe centerline is further accurately fitted using the density clustering algorithm.The method achieves a root mean square error of approximately 0.72 pixels and exhibits a maximum fluctuation of 13.6%under conditions of fewer than 1000 random noise points.In terms of performance,the proposed method demonstrates a 4.8-fold increase in efficiency compared to the conventional Steger algorithm,with a minimal loss of 9.2%in accuracy.