无法用鼠标选择在帧缓冲区中渲染的四边形

Unable to mouse pick a quad rendered in a framebuffer

奋力鼠标挑一个point/quad。我相信我要么在错误的 space 中使用坐标,要么可能没有考虑帧缓冲区的 position/size(它是主 window 的子 window)。

尝试转换为各种不同的坐标 spaces 并反转模型矩阵。当前正在世界 space 中投射一条射线(希望正确)并尝试将其与点的(四边形)位置进行比较。该点在本地 space 中指定,但实体在原点 (0f, 0f, 0f) 处呈现,因此我认为它在世界 space?[=15= 中应该没有任何不同]

获取世界中的鼠标射线space:

private fun calculateRay(): Vector3f {
        val mousePosition = Mouse.getCursorPosition()
        val ndc = toDevice(mousePosition)
        val clip = Vector4f(ndc.x, ndc.y, -1f, 1f)
        val eye = toEye(clip)
        return toWorld(eye)
    }

private fun toDevice(mousePosition: Vector2f): Vector2f {
        mousePosition.x -= fbo.x // Correct thing to do?
        mousePosition.y -= fbo.y
        val x = (2f * mousePosition.x) / fboSize.x - 1
        val y = (2f * mousePosition.y) / fboSize.y - 1
        return Vector2f(x, y)
    }

private fun toEye(clip: Vector4f): Vector4f {
        val invertedProjection = Matrix4f(projectionMatrix).invert()
        val eye = invertedProjection.transform(clip)
        return Vector4f(eye.x, eye.y, -1f, 0f)
    }

    private fun toWorld(eye: Vector4f): Vector3f {
        val viewMatrix = Maths.createViewMatrix(camera)
        val invertedView = Matrix4f(viewMatrix).invert()
        val world = invertedView.transform(eye)
        return Vector3f(world.x, world.y, world.z).normalize()
    }

悬停在点(11.25, -0.75)上时,光线坐标为(0.32847548, 0.05527423)。我尝试标准化点的位置,但仍然不匹配。

感觉我在missing/overlooking什么东西,或者只是错误地操纵了坐标系。任何见解将不胜感激,谢谢。

编辑更多信息:

四边形的顶点是: (-0.5f, 0.5f, -0.5f, -0.5f, 0.5f, 0.5f, 0.5f, -0.5f)

正在将矩阵加载到着色器:

private fun loadMatrices(position: Vector3f, rotation: Float, scale: Float, viewMatrix: Matrix4f, currentRay: Vector3f) {
        val modelMatrix = Matrix4f()
        modelMatrix.translate(position)
        modelMatrix.m00(viewMatrix.m00())
        modelMatrix.m01(viewMatrix.m10())
        modelMatrix.m02(viewMatrix.m20())
        modelMatrix.m10(viewMatrix.m01())
        modelMatrix.m11(viewMatrix.m11())
        modelMatrix.m12(viewMatrix.m21())
        modelMatrix.m20(viewMatrix.m02())
        modelMatrix.m21(viewMatrix.m12())
        modelMatrix.m22(viewMatrix.m22())
        modelMatrix.rotate(Math.toRadians(rotation.toDouble()).toFloat(), Vector3f(0f, 0f, 1f))
        modelMatrix.scale(scale)
        shader.loadModelViewMatrix(viewMatrix.mul(modelMatrix))
        shader.loadProjectionMatrix(projectionMatrix)
    }

在顶点着色器中计算 gl_Position:

gl_Position = projectionMatrix * modelViewMatrix * vec4(position, 0.0, 1.0);

编辑 2:根据 Rabbid 的评论 material 阅读了更多内容后更改了我的代码。不确定我是否需要在视口大小中除以 2(我有视网膜 MacBook 显示器)。

mousePosition.sub(fboPosition)
        val w = (fboSize.x / 2).toInt()
        val h = (fboSize.y / 2).toInt()
        val y = h - mousePosition.y

        val viewMatrix = Maths.createViewMatrix(camera)
        val origin = Vector3f()
        val dir = Vector3f()
        Matrix4f(projectionMatrix).mul(viewMatrix)
                                  .unprojectRay(mousePosition.x, y, intArrayOf(0, 0, w, h), origin, dir)

window 空间的左上原点是 (0,0)。因此,如果你得到 (0, 0),如果鼠标位于 window 的左上角,你必须跳过:

mousePosition.x -= fbo.x // Correct thing to do?
mousePosition.y -= fbo.y

由于framebuffer的左下角是(0,0),y坐标需要翻转:

val y = 1 - (2f * mousePosition.y) / fboSize.y

Cartesian coordinate is transformed by a (inverse) projection matrix, then the result is a Homogeneous coordinates. You've to do a Perspective divide时,获取视图中的笛卡尔坐标space:

val eye = invertedProjection.transform(clip)
return Vector3f(eye.x/eye.w, eye.y/eye.w, eye.z/eye.w)