ORB implementation in Java(Android package)
Hi,
Im searching for a simple implementation of comparison , between two images using java (Android package). Using non-deprecated functions..
Any help would be welcomed..
i'm vaguely guessing: you read a silly blog post, that recommended using ORB features for image comparison.
a doomed idea. feature matching can be used to find the position of a known part of the scene but not for comparing arbitrary images for similarity.
back to the drawing board (aka own research) for you.
I searching for it a while already..
I started with this, and this, and read here, and here , and here . and I have countless "here" 's.
opencv hard to understand (Im using 3.4.3 (Android)), part of the docs is not exists and every example that I find is Deprecated.
I understand that ORB(feature Det.) is the way that I need to take. But now I dont find any example or explanation how to use ORB with Non-Deprecated functions.
short comment : Hi, Thanks on the comment, I worked following this : see here.
I tried to do this
Feature detection is not my solution ?. ( everything I saw by now says it is..) I just searched for any example or explanation how to use ORB with Non-Deprecated functions..(Android package)
don't trust SO, esp. not questions from 2012...
again, you main problem is a poor understanding / description of this part:
that's far too broad / unspecific. what does it mean exactly ? are you trying to find duplicates ? same objects ? similar color schemes ?
you'll only run in circles, until you can clarify this (for your own, to start with)
Its the first step into to build an app that recognize parking signs all over the world.(college project).
As we need to solve small problems for understanding the library (opencv).
We will have a database of images (we will dynamically build it - according the GPS location of the user - because performance reasons ) . --> The user of the app will take a photograph of the sign, the app will need to find the most similar sign (photo) in the database.
We consider to use histograms, but we dont think it will be appropriate because all of the images will be similar.(same colors) Template matching, was also on our mind but, because the sign photograph can be taken in various angles, we thought to try first feature matching.
Part two: In the meantime we go with opencv 2.4.13.6 , because its almost two weeks and we dont find any example on the web in 3.4.*.
For a day progress we have the following code(opencv part): https://pastebin.com/d3MAr2LT
we have some small problems like : Imread need a path and we have a small problem to operate it in emulator of Android studio. and the exact work that happens in the 'Matching' part of the code. But at least we can write and finally starting making first's steps..
p.s : we dont need real time in this point.
of course that the "small problems" is ours to solve.