Opencv 4 Android detect boxes -


i'm developing android app recognizes, store shelf image, boxes (products) present on shelf.

my approach far following:

  1. grayscale
  2. bilateral filter (or gaussianblur, i've found using bilateral filter better preserve edges)
  3. adaptive threshold
  4. dilate (don't know if it's necessary)
  5. canny
  6. findcontours

so, if source image simple, (like b/w drawing of shelf , boxes) can detect them, real images of shelves isn't working.

the main problem single boxes have different "foreground" colors , logos, , steps detect edges of "inner" box (i.e., colours inside edges of box) , gives me totally nosense results. simplicity i'll show belows intermediate results , source image:

1)source source image

2)grayscale grayscaled image

3)bilateral filtering (in case gaussianblur) bilateral

4)adaptive threshold adaptive thresh

5)dilation dilation

6)canny canny

as can see, because cannot remove foreground of each box, edges given logos or text come play , noises results.

how can overcome problem?

my ony idea trying "remove" or decolouring inner boxes, don't konw how it! all!


Comments

Popular posts from this blog

Hatching array of circles in AutoCAD using c# -

ios - UITEXTFIELD InputView Uipicker not working in swift -

Python Pig Latin Translator -