Message Boards Message Boards

0
|
2768 Views
|
1 Reply
|
4 Total Likes
View groups...
Share
Share this post:

Exame mark table detection

Posted 10 years ago

Dear Community

Exists a way in Mathematica to detect marker fields on an image, see attached draft. The possible marker matrix is always the same only the marked position can vary. For example it means according to the attachment, the first marked field in the first row gives the coordinate/position B1.

Many thanks in advanced for your support.

Best regards

Georg

enter image description here

POSTED BY: Joerg Kuehne

Hi,

here is something to start with:

img = Import["~/Desktop/marker00.jpg"]

enter image description here

First we need a grid. If the boxes are more or less in the same position then we can create a template grid; I used the GetCoordinates tool:

coords = {{80.5`, 321.5`}, {78.5`, 260.0}, {79.5`, 182.5`}, {78.5`, 115.5`}, {80.5`, 51.5`}, {145.5`, 322.5`}, {214.5`, 322.5`}, {281.5`, 322.5`}, {350.5`, 324.5`}, {424.5`, 323.5`}};
Show[img, Graphics[Point /@ coords]]

enter image description here

The dots mark my coordinates. This gives me a coordinate grid:

Show[img, 
 Graphics@Line[{{Min[coords[[1 ;; 5]][[All, 1]]], #[[2]]}, {Max[
        coords[[6 ;;]][[All, 1]]], #[[2]]}}] & /@ coords[[1 ;; 5]], 
 Graphics@Line[{{#, Min[coords[[1 ;; 5]][[All, 2]]]}, {#, 
       Max[coords[[1 ;; 5]][[All, 2]]]}}] & /@ 
  Join[{Min[coords[[1 ;; 5]][[All, 1]]]}, coords[[6 ;;]][[All, 1]]]]

enter image description here

I can extract the relevant coordinate for the x- and y-axes:

xaxis = Reverse@Join[{Min[coords[[1 ;; 5]][[All, 1]]]}, coords[[6 ;;]][[All, 1]]];
yaxis = Reverse@coords[[1 ;; 5]][[All, 2]];

which I can then use to cut the respective fields out. To get a clearer representation I plot them in a grid:

Grid[ArrayReshape[
  ImageTrim[img, #] & /@ 
   Reverse@Flatten[
     Table[{{xaxis[[i + 1]], yaxis[[j + 1]]}, {xaxis[[i]], 
        yaxis[[j]]}}, {j, 1, Length[yaxis] - 1}, {i, 1, 
       Length[xaxis] - 1}], 1], {4, 5}], Frame -> All]

enter image description here

Next, we binaries:

Grid[ArrayReshape[
  Binarize /@ (ImageTrim[img, #] & /@ 
     Reverse@Flatten[
       Table[{{xaxis[[i + 1]], yaxis[[j + 1]]}, {xaxis[[i]], 
          yaxis[[j]]}}, {j, 1, Length[yaxis] - 1}, {i, 1, 
         Length[xaxis] - 1}], 1]), {4, 5}], Frame -> All]

and count the pixels:

graydata = 
 ArrayReshape[
  Total[Flatten[
      ImageData[
       ColorNegate@#]]] & /@ (Binarize /@ (ImageTrim[img, #] & /@ 
       Reverse@Flatten[
         Table[{{xaxis[[i + 1]], yaxis[[j + 1]]}, {xaxis[[i]], 
            yaxis[[j]]}}, {j, 1, Length[yaxis] - 1}, {i, 1, 
           Length[xaxis] - 1}], 1])), {4, 5}];

We can plot this to check:

graydata // ArrayPlot

enter image description here

There are obviously two clusters:

clusts = FindClusters[Flatten[graydata], 2];

Which we can use to attach a "1" to "ticked" and a "0" to "unticked".

Flatten[Table[# -> 2. - i & /@ SortBy[clusts, Total][[i]], {i, 1, 2}]]

That gives us the matrix we want:

Grid[graydata /. Flatten[Table[# -> 2. - i & /@ SortBy[clusts, Total][[i]], {i, 1, 2}]], Frame -> All]

enter image description here

Note, that this describes the train of thought. We can wrap everything up into this nice and easily readable function:

tickedBoxes[img_Image] := 
 Module[{}, 
  coords = {{80.5`, 321.5`}, {78.5`, 260.0}, {79.5`, 182.5`}, {78.5`, 
     115.5`}, {80.5`, 51.5`}, {145.5`, 322.5`}, {214.5`, 
     322.5`}, {281.5`, 322.5`}, {350.5`, 324.5`}, {424.5`, 323.5`}};
  yaxis = Reverse@coords[[1 ;; 5]][[All, 2]]; 
  xaxis = Reverse@Join[{Min[coords[[1 ;; 5]][[All, 1]]]}, coords[[6 ;;]][[All, 1]]];
  graydata = ArrayReshape[Total[Flatten[ImageData[ColorNegate@#]]] & /@ (Binarize /@ (ImageTrim[img, #] & /@ 
         Reverse@Flatten[Table[{{xaxis[[i + 1]], yaxis[[j + 1]]}, {xaxis[[i]], yaxis[[j]]}}, {j, 1, Length[yaxis] - 1}, {i, 1, Length[xaxis] - 1}], 1])),{4, 5}]; 
  clusts = FindClusters[Flatten[graydata], 2]; 
  graydata /. Flatten[Table[# -> 2. - i & /@ SortBy[clusts, Total][[i]], {i, 1, 2}]]]

If you throw your image in it should work:

tickedBoxes[img]
(*{{0., 1., 0., 0., 0.}, {0., 0., 1., 0., 0.}, {0., 1., 0., 0., 0.}, {0., 0., 0., 1., 0.}}*)

There is a lot we can improve, but it should be a starting point.

Cheers,

Marco

POSTED BY: Marco Thiel
Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
Attachments
Remove
or Discard

Group Abstract Group Abstract