3-D camera calibration using vanishing point concept

Abstract

A three-dimensional (3-D) camera calibration technique that uses two-dimensional (2-D) grid paper and a vanishing point concept was developed and tested. The calibration technique does not require knowledge of the precise location of the calibration target as other calibration techniques require. The algorithm developed consists of the following steps: (1) Use a point-finder program with a subpixel interpolation technique to identify the points on a grid. (2) Apply an equalization procedure. (3) Use a line-finder program with a statistical linear regression technique to identify a line. (4) Repeat Steps 1 and 2 until all the lines are found. Then divide the lines into two groups—horizontal and vertical. (5) Determine one vanishing point for all the horizontal lines and one vanishing point for all the vertical lines. (6) Using the same vanishing point concept, determine the vanishing points for the planes of sight in space (horizontal and vertical). (7) Determine the camera location and rotational matrix. By using statistical methods, resolution at the subpixel level was achieved. Commercially available components were used in the vision system for this project. The results of the two tests showed that a low-cost vision system can make 3-D coordinate measurements as exact as 5 mils (0.005 inches) in accuracy and 1 mil in repeatability when properly set up and used with this calibration technique.