First a legal notice.  The process described here is patented.  This description is protected by copyright law.  By publishing this description, we have not granted any license to copy or duplicate or perform this process to anyone.  This is a summary of the more interesting parts of the process for your entertainment only. 

The patents are available for reading on the web from a site maintained by an IBM subsiderary, Delphion, expressly for patent searches.  The first link below will take you to that site.  The next three may get you directly to the applicable patents (things always change).

The first step in processing an almost 13 square foot target full of hundreds of holes is to take its picture.  This is done with an ordinary video camera and ... mirrors (no smoke, please). We use the mirrors to convert the aspect ratio of the video image from 4 wide by 3 high (the industry standard for regular TVs, which used to be the standard for movie film too) to 1 wide by 48 high.  By making the actual size of the image 1 inch wide by 48 inches tall, we can get 80 lines of resolution per inch or 3 lines per mm over the entire area of the target.  This is sufficient to detect BB holes present in the target.

With the TV camera and mirrors ready, we attach the target to a revolving drum made of lexan which has a light inside.  We slowly rotate the drum so the camera can see the whole target.  To get roughly the same resolution along the direction of rotation, we take almost a minute (60 seconds x 60 NTSC fields per second) to pass the whole target in front of the mirror assembly aperture.

During that minute a special computer analyzes the bright and dark areas of the picture (holes and not holes) and produces messages which it sends to a regular PC for further processing.  The messages contain 4 bytes each and can be generated at up to 1,000,000 per second.  If you do the math, 10 bits per byte on a serial interface, 4,000,000 bytes per second, you get 40 MegaBits per second.  This is really fast.  The special computer buffers these messages and feeds them to the regular PC as a rate which the regular PC can handle. 

The amount of processing required to take all these messages and find out where the holes are is substantial and is performed in a series of steps, most of which are not synchronized with either the 40 Megabits per second or the more common 19,200 bps which serial ports handle with ease.

First you should image looking at an image through a screen like the ones that keep bugs out of your house.  The screen we use has over 6000 squares per square inch.  The special computer tells the regular PC which ones were lit up.  The regular PC then decides which of the lit up squares are really part of one BB hole.  By processing 10s of thousands of lit up squares, it makes a list of hundreds of BB holes and a few which looked weird.  The weird ones are usually either lopsided or too big to be a single BB hole.  Another program works with an operator to decide what the weird holes really were.

Some of the weird holes are really two or more BB holes in a constellation which the program could not determine precisely.  Six of the weird holes are reference holes.  Before we send the targets out to be shot, we drill 6 1/4 inch holes at precise locations on the paper target.  One is at the center, the other five are near the edges.  Two of the holes are at the upper left hand corner.  These are used to be sure we have the image right-side up and correct left to right when we print the reports.  It is easy for a BB hole right near a reference hole to be confused because there are so many lit up squares.  The operator can quickly determine if more than one BB hole is present or if a BB hole is hiding next to a reference hole.  Sometimes the operator has to look at the target, not just the view through the screen, to see what really happened.

Now that all the lit up squares have been converted to "features" which we care about, we square up the data.  We take all the data we have for the target including the reference holes.  We know where they are supposed to be.  We treat the target data like it is riding on a rubber sheet.  We stretch the data until the reference holes are where they belong, dragging the BB holes along with them.  Once this is done, we are ready to really analyze the shotgun pattern, which is what this is all about anyway.

The first part of the process involves actually finding the center of the pattern.  This turned out to be tougher than you might think.  We needed a method which would handle patterns from tight to very loose.  We also needed to handle patterns which were off center due to problems with the gun or the gunner.  We tried lots of methods for finding the center of the pattern.  Many were more complicated than the one we chose.  However, most produced poor results on "unusual" patterns and none were better for the easy patterns. 

Now that we know where the center of the pattern is, we can fill the area which is off the paper with Phantom BB holes which are appropriate.  For each BB hole on the paper, an invisible line is drawn from that BB hole to the center of the pattern.  That line is then extended through the center until we reach a spot the same distance from the center as the original BB.  If this new spot is on the paper, we just skip it and go on to the next BB hole.  However, if it is off the edge of the paper, we construct a phantom hole at that spot as a reasonable simulation of what might have been there.  It will make the resulting pattern appear to be a bit more symetric that it actually was.  However, the pattern will have the same percentage lines and density areas as if the pattern had landed in the center of the pattern target.

From here on, the process is specific for each report.  The descriptions of each report tell how they are built so it won't be repeated here.  To get the best understanding of how they are created, you'll have to read the Detailed versions of those descriptions.  The links below will take you to those pages.