Android opengl 左/右方向倾斜视锥体问题
我正在编写一个简单的陀螺仪测试应用程序，其中我希望通过将手机倾斜到不同的方向来获得多边形的不同侧面。它类似于头部跟踪渲染技术，但我假设眼睛位置固定，并且只有手机相对于眼睛（相机）移动。
为此，我在 opengl es 中使用了倾斜的平截头体技巧。我可以毫无问题地在顶部/底部/近/远方向上偏移平截头体。真正奇怪的部分是，当我将偏移值除以 2 时，我只能得到左/右方向。有效地“缩小”了平截头体的实际尺寸。
这是工作示例代码，在左/右方向上存在除以 2 的错误：
// xEyePhone is the transform of eye relative to the screen, note the /2.0f for the 3rd and 4th params.
Matrix.frustumM(_ProjMatrix, 0, -ratio-xEyePhone[12]/2.0f, ratio-xEyePhone[12]/2.0f, 1-xEyePhone[13], 1-xEyePhone[13], xEyePhone[14], xEyePhone[14]+10);
Matrix.setLookAtM(_VMatrix, 0, xEyePhone[12], xEyePhone[13], xEyePhone[14],
XEyeCenterWorld[12], xEyeCenterWorld[13], xEyeCenterWorld[14],
XUpVectorWorld[12], xUpVectorWorld[13], xUpVectorWorld[14]);
有人遇到过这个吗？
I am writing a simple gyro test app in which I want to be able different sides of a polygon by tilting the phone to different orientation. It is similar to the head-tracked rendering technique but I assume a fixed eye position and only the phone moves relative to the eye (camera).
To do this I used the skewed frustum trick in opengl es. I could offset the frustum in the top/bottom/near/far directions without problem. The really weird part came in when I only got the left/right directions to work if I divide the offset values by 2. Effectively 'shrinking' the frustum from what it really should be.
Here is the working sample code with the divide-by-2 bug in the left/right direction:
// xEyePhone is the transform of eye relative to the screen, note the /2.0f for the 3rd and 4th params.
Matrix.frustumM(_ProjMatrix, 0, -ratio-xEyePhone[12]/2.0f, ratio-xEyePhone[12]/2.0f, 1-xEyePhone[13], 1-xEyePhone[13], xEyePhone[14], xEyePhone[14]+10);
Matrix.setLookAtM(_VMatrix, 0, xEyePhone[12], xEyePhone[13], xEyePhone[14],
XEyeCenterWorld[12], xEyeCenterWorld[13], xEyeCenterWorld[14],
XUpVectorWorld[12], xUpVectorWorld[13], xUpVectorWorld[14]);
Anyone came across this?
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
您将观察矩阵应用于投影。你不应该这样做。投影变换仅设置“虚拟镜头”的属性。放置相机发生在模型视图转换中。
如果不将这两个部分分开，则会在实施照明时出现问题。
You applying a look-at matrix to the projection. You shouldn't do this. The projection transformation only sets the properties of the "virtual lens". Placing the camera happens in the modelview transformation.
Not keeping those two separate causes problems when implementing lighting.