|
Shape matching
Anonymous |
18 year
|
Hi,
My Lego NXT tracks red objects and picks them up, but I want it to know if it's an object it should be tracking at all.
I was hoping to use shape matching to add extra intelligence to my robot. I have had no success. I tried making a few mock images to test the SHAPE_MATCH module. The image to compare to is 1. Image 3 is matched with 100% confidence, image 4 also (image 1 flipped). But it can't match anything in image 2, even though there are 2 replicates of Image 1. I also noticed that for image 1, the SHAPE_MATCH module sees 44 blobs. Why is this and what is the significance of these blobs?
Thanks,
PT
|
|
|
|
Image Noise
Anonymous |
18 year
|
PT,
You have the right idea but your test images are not binary images. If you look closely at  we thresholded the image to reveal parts of the edges that are a different intensity. When we used the Thresholding->Threshold to clean up the images to pure binary (only black and white) and ensure that all are binary we get a 98% match on 2 of the blobs in image2. Try thresholding to a pure binary image and things should improve.
Also, on your template image be sure to cut it down to just the size of the shape. In that way the "box" display in the shape_match will better outline the shape since it uses the original image size as the box size.
Let us know how it goes!
STeven.
|
|
|
|
Image noise
Anonymous |
18 year
|
|
Thanks for the prompt response. I'll give it a go.
|
|
 |
Shape matching
Pt from Ireland [2 posts] |
18 year
|
I took your advise and it seems to work a lot better.
Thanks
|
|
This Thread
