Project

General

Profile

action #374

Improve openCV matching

Added by aplanas about 9 years ago. Updated about 9 years ago.

Status:
Closed
Priority:
Normal
Assignee:
Category:
Development
Start date:
2013-04-30
Due date:
2013-05-02
% Done:

100%

Estimated time:
8.00 h

Description

The actual matching algorithm can find a pattern inside an image, but when the art, widget or text change a bit, the matching ratio drop bellow the matching threshold.

We can investigate if reducing the amount of information of an image (via binarize or gauss filter) allow better matching ratio.

History

#1 Updated by aplanas about 9 years ago

  • % Done changed from 0 to 90
  • Status changed from New to In Progress

Improve matching performance from 0.78 to 0.95 using Gauss and gray images, and to 0.98 using blur with 7x7 kernel. I propose using the Gauss 5x5 one, is a bit lower matching ration for the test image, but is better for full different images: if I use blur w/ 7x7 kernel, different images become more similar faster, and that is bad for checkneedle().

#2 Updated by aplanas about 9 years ago

  • Status changed from In Progress to Closed

#3 Updated by aplanas about 9 years ago

  • % Done changed from 90 to 100

Also available in: Atom PDF