Adaptive centre extraction method for structured light stripes

Abstract

Abstract. In the vision systems based on structured light, accurate extracting of centre positions of a light stripe represents one of the key points in solving the whole measurement task. Usually, high radiant illumination and objects with abundant surface textures decrease the detection precision. To address these problems, we suggest an adaptive centre-extraction method. First, it improves the image contrast using a novel adaptive threshold-based power transformation. Second, pixel-level centre points are obtained with our adaptive dual-threshold Canny's edge detection method. Finally, subpixel-level centre points are extracted using a Hessian matrix for a limited number of pixels. Our experiments prove robustness and practicability of the method, which can cope with complex surface textures of projected objects and high radiant illumination.

Weld line structured light stripe center extraction approach based on data dimensionality reduction

“…However, because it needs to carry out five times of two-dimensional Gaussian convolution on the image to calculate the normal, the amount of operation is enormous. To solve this problem, Hu et al [13] proposed a approach based on adaptive threshold power transformation, which improved the extraction speed and met the requirements of real-time extraction in the industry. Liu et al [14] proposed a approach for extracting the center of the light stripe based on the Hessian matrix and region growth approach.…”