[IGSTK-Developers] CylinderSpatialObjectRepresentation

[Luis Ibanez]

It is a matter of concern that the CylinderSpatialObjectRepresentation
is being customized for satisfying a very specific use, namely:


             "To make it look like a pointer in a scene"


This is a missuse of a generic geometrical class, and a poor approach
for a toolkit based on Object Oriented Programming.



If we need a "ScenePointer" object, we should introduce a specific class
such as:

                     "ScenePointerSpatialObject"


then customize it with properties (e.g. Transforms) such as having
the axis to be parallel to Z, and the end to be in the point (0,0,0).



      IGSTK Classes should map straight to well defined concepts.



We already discussed this issue with regard to the NeedleSpatialObject,
and it seems that now we are proceeding backwards.



     Luis



-------------------
Julien Jomier wrote:
> Patrick,
> 
>  > I guess David meant the Pointer is pointing towards the -Z direction.
>  > But anyway, only when I remove the '-' sign, that the cylinderObject
>  > will be on the right side.
> 
> No the pointer is pointing towards the Z direction. This shouldn't 
> matter, you should be able to set the CalibrationTransform to satisfy 
> your needs. To me your registration should only try to optimize the 
> PatientTransform and you should set the CalibrationTransform so that 
> it's inline with your tracker.
> 
> 1. The CylinderSpatialObject is mainly a data container when we talk 
> about tracking because the Tracker will modify its transform directly.
> It is not true that it is only when we visualize that the orientation is 
> correct. The orientation is correct when we track the object because the 
> tracker will place the object in the real-world coordinates correctly.
> 
> 2. To me the real principal axis is Z for now and then. The tracker (i.e 
> the CalibrationTransform and the PatientTransform) should take care 
> placing the object in the correct position/orientation.
> 
> 3. I don't think we have to.
> 
> Hope that makes sense,
> 
> Julien
> 
> 
> Patrick Cheng wrote:
> 
>> I guess David meant the Pointer is pointing towards the -Z direction.
>> But anyway, only when I remove the '-' sign, that the cylinderObject 
>> will be on the right side.
>>
>> The PrincipalAxisCalibration is taking the known principal axis of the 
>> Spatial Object(cylinder) and Tracker Tool(pointer) in their own 
>> coordinate system. eg. By default, the cylinder is along it's Y axis, 
>> and pointer is along its Z axis. The PrincipalAxisCalibration then 
>> take the two principal axis, and figure out what rotation we need to 
>> take to align the cylinder with the pointer (in another word, 
>> transform the cylinder's coordinate system to the pointer's coordinate 
>> system).
>>
>> Julien, I have questions about your approach (add an transform to actor)
>>
>> 1. The final CylinderSpatialObject is actually not carrying the 
>> correct transform. Only when we visualize it, we add another transform 
>> to make the actor look right. I believe we should have the 
>> CylinderSpatialObject transform right instead of altering the actor's 
>> transform. suppose we need to do path planing or collision detection, 
>> we are working with the SpatialObject's transform, not the actor's 
>> transform.
>>
>> 2. This will conflict with the PrincipalAxisCalibration, as people get 
>> confused what is the real principal axis is. Here the principal axis 
>> for the cylinder is still Y, and Pointer is Z. But since we are going 
>> to have a rotation when we visualize it, we don't need the 
>> PrincipalAxisCalibration here.
>>
>> 3. maybe we can move the rotation and shift of origin in the 
>> CylinderSpatialObject class.
>>
>> Patrick
>>
>> Julien Jomier wrote:
>>
>>> Hi Patrick,
>>>
>>> David told me that the cylinder should be located in the [-Z,0], 
>>> therefore the cylinder should be translated by 
>>> -m_CylinderSpatialObject->GetHeight()/2.0
>>>
>>>  > One quick question: If we can just change the transform when we 
>>> create
>>>  > the actor, what do we need the PrincipalAxisCalibration for?
>>>
>>> I don't know what the PrincipalAxisCalibration class is actually 
>>> doing, but the user cannot change the actor, so we need a way to 
>>> calibrate the tool anyway.
>>>
>>> Julien
>>>
>>> Patrick Cheng wrote:
>>>
>>>> Hi Julien,
>>>>
>>>> I checked out your recent change in the 
>>>> CylinderSpatialObjectRepresentation class. Now without setting the 
>>>> rotation part of the CalibrationTransform, the orientation of the 
>>>> CylinderObject will be correct, but with one little bug, in line 152 
>>>> (igstkCylinderObjectRepresentation.cxx)
>>>> cylinderActor->SetPosition(0,0,-m_CylinderSpatialObject->GetHeight()/2.0); 
>>>>
>>>> There should be no '-' sign. I tested it in the application.
>>>>
>>>> One quick question: If we can just change the transform when we 
>>>> create the actor, what do we need the PrincipalAxisCalibration for?
>>>>
>>>> Patrick
>>>>
>>>
>>>
>>
>>
> _______________________________________________
> IGSTK-Developers mailing list
> IGSTK-Developers at public.kitware.com
> http://public.kitware.com/cgi-bin/mailman/listinfo/igstk-developers
> 
>