<br>Hi Ruisheng,<br><br>Unfortunately we don't have specific examples for this<br>configuration of registration (Lidar-to-Image) in the <br>ITK source tree.<br><br>However,<br><br>You may find useful to look at the code of the test:<br>
<br> Insight/Testing/Code/Algorithms/<br><div id=":8g" class="ii gt"> itkPointSetToImageRegistrationTest_1.cxx<br><br><br> Regards,<br><br><br> Luis</div><br>---------------------------------------------------------------<br>
<div class="gmail_quote">On Mon, Apr 12, 2010 at 3:51 PM, ruisheng wang <span dir="ltr"><<a href="mailto:wangruisheng@gmail.com">wangruisheng@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
Hi Luis,<br><br>Thank you very much for your replies. Do you have an
example code showing the process of Lidar-to-Image registration using<br>Mutual
informaiton. I could not find examples for doing this. In the ITK
document, there is not an example showing<br>
PointSetToimageRegistration, do you have an example code showing how to
use PointSetToImageregistration?<br><br>thanks a lot!<br>Ruisheng<br><br><br><br>Hi Ruisheng,<br>
<br>
In order to use this class effectively, you should<br>
first become very familiar with the ITK Registration<br>
Framework.<br>
<br>
Please read the ITK Software Guide:<br>
<br>
<a href="http://www.itk.org/ItkSoftwareGuide.pdf" target="_blank">http://www.itk.org/ItkSoftwareGuide.pdf</a><br>
<br>
In particular the Chapter 8 "Image Registration"<br>
in pdf-pages: 347-533.<br>
<br>
<br>
Once you are familiar with the framework,<br>
you may want use the following configuration:<br>
<br>
<br>
1) Use the PointSetToImageRegistration class<br>
<br>
2) Use the itkPointSetToImageMetric class<br>
<br>
3) Use the perspective transform<br>
<br>
4) Start with an optimizer that does not require<br>
derivatives (e.g. the Amoeba optimizer).<br>
<br>
<br>
Key factors for a successful registration are:<br>
<br>
A) Good initialization of the Transform<br>
<br>
B) Steady setup for the optimizer parameters<br>
<br>
C) Proper scaling of the Transform parameters.<br>
<br>
In your case, please note that (A) includes to<br>
make a very good initialization for the center<br>
of projection and the focal distance, and that<br>
(B) means that you should start with optimizer<br>
settings that barely modify the Transform, and<br>
from them you can progress towards having<br>
the Transform explore a wider range of<br>
parameters.<br>
<br>
<br>
Regards,<br>
<br>
<br>
Luis<br><br><br>Hi everyone,<br><br>I'm trying to register 3D lidar with 2D optical
images using the transform ITK::<div>Rigid3DPerspectiveTransform.<br>But
seems there is not much explanation or examples from ITK to describe
the this functions. I wonder if<br>
anyone here could give me some hints for how to use ITK::Rigid3DPerspectiveTransform
for the 3D-to2D registration<br>problems.<br><br>Thanks very much!<br><font color="#888888">Ruisheng <pre><br></pre></font></div><br>
</blockquote></div><br>