Image Processing in Vision 3D Coordinate Measurement System

In the image processing of a vision three-dimensional (3D) coordinate measurement system which is consisted of a CCD camera, a computer and a special light pen with several LEDs, unknowable noise images hold the way to get the image of LEDs out of the original image in data processing of the system. It is intolerant in real-time measurement system to use a usual method of recognizing object which can give a correct answer only in weak noise background. In this paper, a method using the principle of subtraction is introduced to efficiently separate the image of LEDs from the original image at the first step of the image processing. The positions of LEDs image in original image can easily be estimated when background is subtracted from the original image. The coordinates of image center of these LEDs will be obtained with ellipse fitting methods to analyze some small areas of the original image around the estimated image of LEDs. It theoretically costs computer less half time to use the proposed method than to use the usual method of recognizing object in algorithmic analysis. The experiments show that this method is not only fast but precise in estimating and ascertaining the coordinates of the image of LEDs in the original image.