How to obtain centroids from cascade object detector?

I am having trouble determining a way to get centroids from objects detected by a trained object detector- the only output seems to be bounding boxes. Is there another way I could get the centroid coordinates for the detected HOG features?

code:

detector = vision.CascadeObjectDetector('Detector3.xml');
detector.MergeThreshold = 3;
img = imread('Test_image.jpg');
[bbox] = step(detector,img)

Answer 1

[Image Analyst]

If you have a binary image you could use regionprops(). See my Image Segmentation Tutorial http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862 for a full demo.

[William Maurer]

Thank you, that demo was very helpful!

Answer 2

[Dima Lisin]

Hi William,

The best you can do, is to get the centroid of the box. The bounding box format is [x, y, widht, height]. So the centroid is [x+width/2, y+height/2]. or for multiple bboxes: [bbox(:,1)+bbox(:,3)/2, bbox(:,2)+bbox(:,4)/2].