search

PatrickLutz11 / SoftwareEngineeringProject

Group B: Jannis, David, Patrick
2 stars 0 forks source link

Shape detection with camera #26

Open 99Jannis opened 6 hours ago

99Jannis commented 6 hours ago

if I currently use the Gui with shape detection it outputs this image: image I will try to further look into the source of the problem, but would be glad with some help

99Jannis commented 6 hours ago

problem is probably with shape detection. I encounter the problem with test_image_03.jpg image

99Jannis commented 6 hours ago

result.JPG as well and I think I also found the probable error source. further info in the next comment.

Input result.jpg: image Output result.jpg: image

99Jannis commented 6 hours ago

The shape detection looks for ALL contours of an image and if the blur matrix isn't big enough it detects MORE contours.

So there are probably hundreds of contours it detects and then also wants to detect their color and write it on the image.

99Jannis commented 6 hours ago

possible solution

a relative blur in regards to the size of the image instead of an absolute one, could maybe fix the problem but I'm not sure...

Master-D1990 commented 6 hours ago

We could probably implement some kind of thresholds which are dynamic, reduced until we get a set amount of detected shapes. The downside of it is that it'll almost any time give us the wrong amount of shapes.

Another feature could be a slider in the GUI with the function to set the thresholds manually.

Thus, I'm not sure how important this is for the grading, but we could definitely put some work in it.

99Jannis commented 4 hours ago

possible solution

a relative blur in regards to the size of the image instead of an absolute one, could maybe fix the problem but I'm not sure...

I tried this and it didn't work. It made it even worse. but what improved a bit was the recognition if I changed the blur from gaussian to the non-weighted one