The main scan step is to capture images of an object illuminated by a projected pattern consisting of vertical stripes of different widths. This structured light method allows us to generate a set of images as shown below.

 
Our scan script automatically projects normal and inverse images of the pattern on the object and then the images are captured by our system and stored with specific names in a folder containing all the object's data (for example, pictures and calibration files). The script also projects and captures two additional images: one totally white and another completely black. These images are used to identify shadow areas, to determine a threshold for stripe identification and to create a "mask" image for background removal.

Stripe determination

To find the coordinates of the points at the stripe boundaries in all resolutions we subtract the normal and the negative images of the object and determine with subpixel accuracy where the sign changes in pixels intensity occurs. The zero-crossings represent the boundary points.

Then, we need to identify every stripe with an unique code. Since we projected stripes with different widths using the Graycode pattern, all we need to do is to look at each pixel and give it a code of 1 if it is white or 0 if it is black. Therefore we obtain a N-bit Graycode (where N is the resolution level, for example, 8). Each stripe code is converted to a decimal number representing the x-coordinate in the projector frame of reference. All this procedure is done by our active triangulation module. After that, the program saves a list with the valid boundary points index (those not removed by the mask) and their subpixel coordinate.

Last update December 12, 2003