Finding Shapes in image

Question

Finding Shapes in image
Hi,
I have to determine the location and size of a black rectangle (stroke 1-2 millimeter, black) in a tiff-image.
My current approach is:
1) Start detection in defined loacation (to limit detect-area)
2) Start detection in samples (e.g. 5 per centimeter)
3) evaluate iterative the rgb of pixel and calculate the percental part of color to detect
Does anybody know some Objects providing a convinient way to realize this demand.
imagined code - just a draft:
private void detect(BufferedImage someImage, Rectangle cutRectangle, float expectedPercentage, int detectedRGB){
     //..doing some calcuation to find cut-in points - values defined by Configuration
     Point [] cutinPointsX = getCutInPoints("x");
     Point [] cutinPointsY = getCutInPoints("y");
     float percentageOfRGB;
     // find x
     for(int i = 0, i < cutinPointsX.length; i++){
          percentageOfRGB = getRgbFromRectangle(cutinPointsX, cutRectangle, someImage, detectedRGB);
          if(percentageOfRGB == expectedPercentage){
               // calculate x of DetectedRectangle
     // find y
     for(int i = 0, i < cutinPointsY.length; i++){
          percentageOfRGB = getRgbFromRectangle(cutinPointsY, cutRectangle, someImage, detectedRGB);
          if(percentageOfRGB == expectedPercentage){
               // calculate y of DetectedRectangle
private float getRgbFromRectangle(Point startPoint, Rectangle cutRectangle, BufferedImage someImage, int detectedRGB){
     float rgbPercentage;
     int startx = startPoint.getX();
     int startY = startPoint.getY();
     int xLength = startx + cutRectangle.getWidth();
     int yLength = startY + cutRectangle.getHeigth();
     int rgb;
     for(int x = startX; x < xLength; x++){
          for(int y = 0; y  < yLength; y++){
               currentRGB = someImage.getRGB(x,y);
               if(currentRGB = detectedRGB){
                    rgb++;
     // some calculations
     return rgbPercentage = calculateRGBPercentage(rgb, (xLength * yLength));
...

Answer 1

Hi,
do u mean recognize a shape in a 2d bitmap?
If so I think is has little to do with j3d but it is more like image segmntation problem and data clustering.
I worked on kvm and cvm algos when i was at the university, if this is your problem le me know...
I think u could start looking for object recognizing algos for mobile robots, but I'm afraid it is an hard job.
If the shape must exactly match the prototype (like a square or a star) u could use a fuzzy matching among translation and rotation params, but this hardly cover overlapping images case.
Anyway.. good luck
Mario