Error Distribution of Color-Encoded Projector

摘要

A structured light projector, especially pattern projector, is being applied more and more widely in machine vision. The projector can produce a plenty of optical interest points on the surface of three-dimensional object, which makes its lens quality very important. Otherwise, there will be great error between ideal locations and factual locations of interest points. In the paper, the distortion models of a color-encoded projector's lens are described in details. Based on two-step calibration algorithm, experiments are implemented. The projector is regarded as a reverse camera whose parameters can be calibrated according to point-to-point method, at the same time, the lens distortion must be thought to reform future three-dimensional reconstruction. So, its lens distortion is presented and analyzed after careful calibration. According to the experiment result, the paper suggests that central area of a projector lens is more appropriate than its marginal area in active vision system and equivalent focal length should be enlarge in order to decrease the influence of lens distortion. When the error from lens distortion is too much to accept, error compensation of encoded interest points in pattern should be carried out. It is worthwhile noting that the location of all the interest points is fixed and the most attentive information is just their locations during the reconstruction algorithm. Thereby, once the distortion is gained, it is considered as a kind of principal systemic error and it can be removed directly in computer program for color codes.