Re: [osg-users] Camera rotation on three axes
- From: Paul Melis at 2008-11-05 20:06- Hi Steven, Maybe the following is an easier way to accomplish what you want, as I don't see why you are trying to calculate angles... snip Hi Paul, Thanks for all your help. I've finally tried enough methods to appreciate the differences and how I can avoid gimbal lock problems. I'm now using pitch to select the latitude in my sphere of view rather than an oblique great circle. This seems to work well for my purposes. Thanks again, Steven ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Camera rotation on three axes
- From: John Vidar Larring at 2008-10-30 21:42- Hi Steven, Here is a snipplet of code from the osgautocapture example that I just sent to osg-submissions (if my email got through?), provided as is. I hope it might be of help: Hi John, Thanks for the code. I tried it (and it also rekindled my interest in ViewMatrixLookAt) but I still got gimbal-lock. Perhaps my implementation of your code is faulty. I've spent days now trying to learn enough about quaternions. This almost works but my calculation for the rotation angle is still wrong. Is there a function in OSG for calculating a quaternion from euler angles ? I've searched but haven't found anything. The standard code to calculate a quaternion uses w = cosx * cosy * cosz + sinx * siny * sinz. This doesn't seem right to me and it doesn't work in my testing. I would expect the final rotation to be between the largest input rotation and PI. I hope this post is sufficiently on-topic for the osg-users list. I apologise if I should be looking elsewhere. Thanks, Steven ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Camera rotation on three axes
Hi, Steven Saunderson wrote: - From: John Vidar Larring at 2008-10-30 21:42- Hi Steven, Here is a snipplet of code from the osgautocapture example that I just sent to osg-submissions (if my email got through?), provided as is. I hope it might be of help: Hi John, Thanks for the code. I tried it (and it also rekindled my interest in ViewMatrixLookAt) but I still got gimbal-lock. Perhaps my implementation of your code is faulty. I've spent days now trying to learn enough about quaternions. This almost works but my calculation for the rotation angle is still wrong. Is there a function in OSG for calculating a quaternion from euler angles ? I've searched but haven't found anything. You can make a quat from three angles and three axes: osg::Quat lori = osg::Quat(osg::DegreesToRadians(heading), osg::Vec3d(0,0,1), osg::DegreesToRadians(pitch), osg::Vec3d(0,1,0), osg::DegreesToRadians(roll), osg::Vec3d(1,0,0)); Make sure the order is what you need. The standard code to calculate a quaternion uses w = cosx * cosy * cosz + sinx * siny * sinz. This doesn't seem right to me and it doesn't work in my testing. I would expect the final rotation to be between the largest input rotation and PI. I hope this post is sufficiently on-topic for the osg-users list. I apologise if I should be looking elsewhere. Thanks, Steven jp -- This message is subject to the CSIR's copyright terms and conditions, e-mail legal notice, and implemented Open Document Format (ODF) standard. The full disclaimer details can be found at http://www.csir.co.za/disclaimer.html. This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean. MailScanner thanks Transtec Computers for their support. ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Camera rotation on three axes
Hi Steven, Steven Saunderson wrote: - From: John Vidar Larring at 2008-10-30 21:42- Here is a snipplet of code from the osgautocapture example that I just sent to osg-submissions (if my email got through?), provided as is. I hope it might be of help: Thanks for the code. I tried it (and it also rekindled my interest in ViewMatrixLookAt) but I still got gimbal-lock. Perhaps my implementation of your code is faulty. I've spent days now trying to learn enough about quaternions. This almost works but my calculation for the rotation angle is still wrong. Is there a function in OSG for calculating a quaternion from euler angles ? I've searched but haven't found anything. Maybe the following is an easier way to accomplish what you want, as I don't see why you are trying to calculate angles... Assume that you want to apply the pitch, roll and yaw rotations to your plane in that order. Also assume that you have your plane centered at the origin with its nose looking down the positive Y axis and its top-side pointing in direction of the positive Z axis. The pitch rotation can then simply be around the global X axis. However, the roll and yaw rotations will need to be around axes other than the global Y and Z, as otherwise they don't take the changed plane orientation due to pitch (and roll for the yaw rotation) into account. For this, keep track of two helper points (osg::Vec3's): the first one, p1, is at (0, 1, 0), the second one, p2, is at (0, 0, 1). When you have calculated your pitch transformation, using for example, osg::Quat(pitch_angle /* in radians */, osg::X_AXIS), you also transform p1 with this quaternion. Let's call this transformed point p1t, and you can simply p1t = q * p1 to get the transformed point. To apply the roll rotation to the plane we then need to rotate around the vector that goes through pt1 and the origin. This vector is simply p1t. So the roll rotation would then be osg::Quat(roll_angle, p1t). To apply the yaw rotation we need to transform p2 first by the pitch rotation followed by the roll rotation: p2t = pitch_transform * roll_transform * p2. Then the yaw rotation will be osg::Quat(yaw_angle, p2t). We now have three rotations defined by three quaternions. We only have to combine these into a single transformtion to get the final plane transformation: q_final = q_pitch * q_roll * q_yaw To convert q_final to a transformation matrix that you can pass to e.g. osg::MatrixTransform::setMatrix() you can use osg::Matrix m; q_final.get(m); // m will now contain the quaternion in matrix form Hope this helps, Paul ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Camera rotation on three axes
Steven Saunderson wrote: I'm trying to rotate a camera on all axes but my code fails when the x-axis or y-axis rotation approaches 90 degrees. I am using a viewer.frame loop to avoid the default manipulator supplied by viewer.run. My current camera code is : cameraRotation.makeRotate ( yrads,osg::Vec3 (0, 1, 0) , // roll Y xrads,osg::Vec3 (1, 0, 0) , // pitch X zrads,osg::Vec3 (0, 0, 1) ); // yaw Z cameraTrans.makeTranslate (xdist, ydist, zdist); myCameraMatrix = cameraRotation * cameraTrans; inverseMatrix = myCameraMatrix.inverse (myCameraMatrix); inverseMatrix *= osg::Matrixd::rotate (-(osg::PI_2), 1, 0, 0); viewer.getCamera ()-setViewMatrix (inverseMatrix); When the pitch (xrads) is 0 everything works properly. But when xrads is PI/2 the effect of yrads is identical to zrads (both do z-axis rotation). I've tried applying the rotations in various orders but the code fails when one of the rotations is PI/2. Can anyone suggest what I am doing wrong here or point me to a sample of working code ? You've hit upon a phenomenon called 'gimbal lock'. See http://web.archive.org/web/20041029003853/http:/www.j3d.org/matrix_faq/matrfaq_latest.html#Q34 for an explanation (and possibly some different ways of controling your camera). Paul ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Camera rotation on three axes
As Paul mentions the is the age old Gimbal Lock problem The quick answer to this is to build a quaternion and use that populate the oration matrix Or the really old nasty trick of limiting rotations to -89.5 thru +89.5 or so ... Gordon __ Gordon Tomlinson Product Manager 3D Email : gtomlinson @ overwatch.textron.com __ (C): (+1) 571-265-2612 (W): (+1) 703-437-7651 Self defence is not a function of learning tricks but is a function of how quickly and intensely one can arouse one's instinct for survival - Master Tambo Tetsura -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Paul Melis Sent: Thursday, October 30, 2008 6:10 AM To: OpenSceneGraph Users Subject: Re: [osg-users] Camera rotation on three axes Steven Saunderson wrote: I'm trying to rotate a camera on all axes but my code fails when the x-axis or y-axis rotation approaches 90 degrees. I am using a viewer.frame loop to avoid the default manipulator supplied by viewer.run. My current camera code is : cameraRotation.makeRotate ( yrads,osg::Vec3 (0, 1, 0) , // roll Y xrads,osg::Vec3 (1, 0, 0) , // pitch X zrads,osg::Vec3 (0, 0, 1) ); // yaw Z cameraTrans.makeTranslate (xdist, ydist, zdist); myCameraMatrix = cameraRotation * cameraTrans; inverseMatrix = myCameraMatrix.inverse (myCameraMatrix); inverseMatrix *= osg::Matrixd::rotate (-(osg::PI_2), 1, 0, 0); viewer.getCamera ()-setViewMatrix (inverseMatrix); When the pitch (xrads) is 0 everything works properly. But when xrads is PI/2 the effect of yrads is identical to zrads (both do z-axis rotation). I've tried applying the rotations in various orders but the code fails when one of the rotations is PI/2. Can anyone suggest what I am doing wrong here or point me to a sample of working code ? You've hit upon a phenomenon called 'gimbal lock'. See http://web.archive.org/web/20041029003853/http:/www.j3d.org/matrix_faq/m atrfaq_latest.html#Q34 for an explanation (and possibly some different ways of controling your camera). Paul ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.or g ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Camera rotation on three axes
Hi Steven, Here is a snipplet of code from the osgautocapture example that I just sent to osg-submissions (if my email got through?), provided as is. I hope it might be of help: ... // Compute eye point in world coordiantes osg::Vec3d eye; csn-getEllipsoidModel()-convertLatLongHeightToXYZ(lat, lon, alt, eye.x(), eye.y(), eye.z()); // Build matrix for computing target vector osg::Matrixd target_matrix = osg::Matrixd::rotate(-heading, osg::Vec3d(1,0,0), -lat, osg::Vec3d(0,1,0), lon, osg::Vec3d(0,0,1)); // Compute tangent vector ... osg::Vec3d tangent = target_matrix.preMult(osg::Vec3d(0, 0, 1)); // Compute non-inclined, non-rolled up vector ... osg::Vec3d up(eye); up.normalize(); // Incline by rotating the target- and up vector around the tangent/up-vector // cross-product ... osg::Vec3d up_cross_tangent = up ^ tangent; osg::Matrixd incline_matrix = osg::Matrixd::rotate(incline, up_cross_tangent); osg::Vec3d target = incline_matrix.preMult(tangent); // Roll by rotating the up vector around the target vector ... osg::Matrixd roll_matrix = incline_matrix * osg::Matrixd::rotate(roll, target); up = roll_matrix.preMult(up); viewerWindow-getCamera()-setViewMatrixAsLookAt(eye, eye+target, up); ... Best regards, John Steven Saunderson wrote: I'm trying to rotate a camera on all axes but my code fails when the x-axis or y-axis rotation approaches 90 degrees. I am using a viewer.frame loop to avoid the default manipulator supplied by viewer.run. My current camera code is : cameraRotation.makeRotate ( yrads,osg::Vec3 (0, 1, 0) , // roll Y xrads,osg::Vec3 (1, 0, 0) , // pitch X zrads,osg::Vec3 (0, 0, 1) ); // yaw Z cameraTrans.makeTranslate (xdist, ydist, zdist); myCameraMatrix = cameraRotation * cameraTrans; inverseMatrix = myCameraMatrix.inverse (myCameraMatrix); inverseMatrix *= osg::Matrixd::rotate (-(osg::PI_2), 1, 0, 0); viewer.getCamera ()-setViewMatrix (inverseMatrix); When the pitch (xrads) is 0 everything works properly. But when xrads is PI/2 the effect of yrads is identical to zrads (both do z-axis rotation). I've tried applying the rotations in various orders but the code fails when one of the rotations is PI/2. Can anyone suggest what I am doing wrong here or point me to a sample of working code ? Thanks, Steven ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org -- Best regards, John WeatherOne -- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean. ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Camera rotation on three axes
- From: Paul Melis at 2008-10-30 21:10- You've hit upon a phenomenon called 'gimbal lock'. See http://web.archive.org/web/20041029003853/http:/www.j3d.org/matrix_faq/matrfaq_latest.html#Q34 for an explanation (and possibly some different ways of controling your camera). Hi Paul, Thanks for you quick answer. I've tried loading the matrix manually but I still get stuck when both x-axis and y-axis are PI/2. I am doing a mult after to apply the x,y,z offsets and then getting the inverse matrix before loading the camera. Will these actions thwart the fix ? Thanks, Steven ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Camera rotation on three axes
- From: Paul Melis at 2008-10-30 23:45- Steven Saunderson wrote: Hi Paul, Thanks for you quick answer. I've tried loading the matrix manually but I still get stuck when both x-axis and y-axis are PI/2. I'm not sure what you mean here. It sounds like you're building your matrix from Euler angles manually, but this will still not solve your gimbal lock problem. You're quite right here. I'm now reading the entire FAQ especially about quaternions. To quote Q39 in the link above: Don't let the spectre of Gimbal Lock fool you: Euler angles are still a complete representation of any rotation in 3D space; it's just that the actual Euler angles needed to achieve some particular desired rotation may be rather unintuitive. Does this mean that instead of rotating both x and y 90 degrees I should find another axis and rotate around this ? What exactly are you trying to achieve? Where do the pitch, rol, yaw values come from, are they user-specified? It is a flight simulator. The pitch, roll, and yaw values come from an external source so any combination is possible. Thanks for the FAQ link; it is a great resource. -- Steven ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
