Novel sub-pixel feature point extracting algorithm for three-dimensional measurement system with linear-structure light

Light strip feature point extracting is a key topic in three-dimensional measurement system with linear-structure light developing process. A novel algorithm of sub-pixel feature point extracting is proposed. The algorithm uses Gaussian low-pass filter to smooth light strip image, adopts edge detection and morphological algorithm to obtain segmentation regions of light strip image. According to it's curvature properties, edge curve of a segmented region are divided into two data sets. K-d tree algorithm is used to search for the closest point from one data set to another and build up two points group. Based on each points group, image gray distribution of a light strip section can be obtained, and then the feature point of one light strip section is calculated with spline interpolation algorithm. The position of extracting feature point in light strip image is selected flexibly on the basis of measurement system parameters. Therefore, the algorithm can effectively correct feature point extracting error for light strip image with asymmetric and incomplete property. The results show the algorithm not only extracts the strip feature point accurately, and but also has good anti-noise and robust performance and more extensive applicability.