Home Messages Index
[Date Prev][Date Next][Thread Prev][Thread Next]
Author IndexDate IndexThread Index

Re: doorplate recognition

  • Subject: Re: doorplate recognition
  • From: Roy Schestowitz <newsgroups@xxxxxxxxxxxxxxx>
  • Date: Mon, 22 Aug 2005 07:59:20 +0100
  • Newsgroups: sci.image.processing
  • Organization: schestowitz.com / Manchester University
  • References: <1124415374.630550.193060@f14g2000cwb.googlegroups.com> <de3rno$27vo$1@godfrey.mcc.ac.uk> <1124675748.742442.278340@g14g2000cwa.googlegroups.com>
  • Reply-to: newsgroups@xxxxxxxxxxxxxxx
  • User-agent: KNode/0.7.2
On Monday 22 August 2005 02:55, zrbj wrote:

> Thank you for your help. Here are some links of the images
> http://www.ia.ac.cn/personal/rui.zheng/doorplate1.jpg
> http://www.ia.ac.cn/personal/rui.zheng/doorplate2.jpg
> http://www.ia.ac.cn/personal/rui.zheng/doorplate3.jpg
> Actually, this program is used for robot, so  photo plane is not
> usually parallel to door,and the doorplate is not like a rectangle but
> parallelogram.I think it's diffcult to build color model to distinguish
> door and doorplate due to the acute variation of illumination
> condition, and i want robot to do the recognation automatically, get
> rid of color calibing every time.

The four metal pieces (similar to large bolts) appear quite clearly in these
images and form a parallelogram. Perhaps you can outline the doorplate by
locating one such metal piece and then stretching lines in all 360 degrees
to find matching points. To gain confidence, you can check the angles
between all of these lines that were said to be a match (angles should sum
up to 360 and there is geometry to muck about with).

The metal does not appear to change much in terms of intensity due to
illumination. This means that once you find one, you have some expectation
as for the colour of the other three.


Roy S. Schestowitz        Useless fact: 85% of plant life in in the oceans

[Date Prev][Date Next][Thread Prev][Thread Next]
Author IndexDate IndexThread Index