正文
WorldWind源码剖析系列：WorldWind如何确定与视点相关的地形数据的LOD层级与范围
提示：扫一扫查出行【扫一扫了解最新限行尾号】
复制提示
1 、 WorldWind如何确定与视点相关的地形数据的LOD层级与范围？
问题描述 ：WW中是如何判断LOD层次的呢，即在什么情况下获得哪一层级的数据？是否只通过相机视点的高度进行判断？
问题切入 ：要解决这个问题，我先说明一下WW的渲染机制，在渲染线程中，Render函数只负责渲染可渲染物体，而不负责视点的更新和Lod的判断。在m_World.Render(this.drawArgs)中可渲染的物体都是通过另一个更新线程WorkerThreadFunc来控制的，具体由m_World.Update(this.drawArgs)实现。因此要找到这个问题的答案需要从Update着手。重点注意m_World.Update(this.drawArgs)中的两段代码。如下所示：
可渲染物体的递归更新：
if (this.RenderableObjects != null)
{
this.RenderableObjects.Update(drawArgs);
}
if (this.m_WorldSurfaceRenderer != null)
{
this.m_WorldSurfaceRenderer.Update(drawArgs);
}
if (this.m_projectedVectorRenderer != null)
{
this.m_projectedVectorRenderer.Update(drawArgs);
}
相机视点的高度更新：
if (this.TerrainAccessor != null)
{
if (drawArgs.WorldCamera.Altitude < )
{
if (System.DateTime.Now - this.lastElevationUpdate > TimeSpan.FromMilliseconds())
{
drawArgs.WorldCamera.TerrainElevation = (short)this.TerrainAccessor.GetElevationAt(drawArgs.WorldCamera.Latitude.Degrees, drawArgs.WorldCamera.Longitude.Degrees, 100.0 / drawArgs.WorldCamera.ViewRange.Degrees);
this.lastElevationUpdate = System.DateTime.Now;
}
}
else
drawArgs.WorldCamera.TerrainElevation = ;
}
else
{
drawArgs.WorldCamera.TerrainElevation = ;
}
上面代码是WorldWind V1.4.0.0中实现的。但是在WorldWind V1.4.0.1中被移入worldwindow.cs文件的WorldWindow.Render()函数中。如下图选中的代码。
问题答案 ：搜索全局Update，重点关注SurfaceTile的Update和QuadTile的Update。
SurfaceTile的Update：
if(drawArgs.WorldCamera.TrueViewRange < Angle.FromDegrees(3.0f*tileSize) && MathEngine.SphericalDistance( Angle.FromDegrees(centerLatitude), Angle.FromDegrees(centerLongitude), drawArgs.WorldCamera.Latitude, drawArgs.WorldCamera.Longitude) < Angle.FromDegrees( 2.9f*tileSize )& drawArgs.WorldCamera.ViewFrustum.Intersects(m_BoundingBox))
{
if(m_NorthWestChild == null || m_NorthEastChild == null || m_SouthWestChild == null || m_SouthEastChild == null)
{
ComputeChildrenTiles(drawArgs);
}
else
{
if(m_NorthEastChild != null)
{
m_NorthEastChild.Update(drawArgs);
}
if(m_NorthWestChild != null)
{
m_NorthWestChild.Update(drawArgs);
}
if(m_SouthEastChild != null)
{
m_SouthEastChild.Update(drawArgs);
}
if(m_SouthWestChild != null)
{
m_SouthWestChild.Update(drawArgs);
}
}
}
else
{
if(m_NorthWestChild != null)
{
m_NorthWestChild.Dispose();
m_NorthWestChild = null;
} if(m_NorthEastChild != null)
{
m_NorthEastChild.Dispose();
m_NorthEastChild = null;
} if(m_SouthEastChild != null)
{
m_SouthEastChild.Dispose();
m_SouthEastChild = null;
} if(m_SouthWestChild != null)
{
m_SouthWestChild.Dispose();
m_SouthWestChild = null;
}
}
QuadTile的Update：
if (DrawArgs.Camera.ViewRange < Angle.FromDegrees(QuadTileSet.TileDrawDistance * tileSize)&& MathEngine.SphericalDistance(CenterLatitude, CenterLongitude,DrawArgs.Camera.Latitude, DrawArgs.Camera.Longitude) < Angle.FromDegrees(QuadTileSet.TileDrawSpread * tileSize)&& DrawArgs.Camera.ViewFrustum.Intersects(BoundingBox))
{
if (northEastChild == null || northWestChild == null || southEastChild == null || southWestChild == null)
{
ComputeChildren(drawArgs);
} if (northEastChild != null)
{
northEastChild.Update(drawArgs);
} if (northWestChild != null)
{
northWestChild.Update(drawArgs);
} if (southEastChild != null)
{
southEastChild.Update(drawArgs);
} if (southWestChild != null)
{
southWestChild.Update(drawArgs);
}
}
else
{
if (northWestChild != null)
{
northWestChild.Dispose();
northWestChild = null;
} if (northEastChild != null)
{
northEastChild.Dispose();
northEastChild = null;
} if (southEastChild != null)
{
southEastChild.Dispose();
southEastChild = null;
} if (southWestChild != null)
{
southWestChild.Dispose();
southWestChild = null;
}
}
在这里我们看到判断Lod的级别主要有三个条件：
1、相机视角范围，视角范围越大，所包含的tileSize就越大；
2、相机与瓦片距离，距离越远，所包含的tileSize也就越大；
3、相机视锥与瓦片是否相交。
相对应我们可以把视角剔除方法理解成以下三步：
1、根据视角范围，画个大圈，把大圈里的大瓦片全加进来；
2、根据相机与瓦片的距离进行细化瓦片，如果太远的瓦片仍然保持很大，而近处的瓦片需要进一步更新，计算子瓦片；
3、每计算一个瓦片都需要判断它是否在视锥里，否则剔除身后看不到的瓦片。
在理解以上三个条件的前提下，需要理解的是视角、距离到底与tileSize有什么关系。对于不同级别影像或地形，ww中设置了相应的大小，根据一定的比例系数和经验参数寻找与距离和视角的关系。即通过调整参数 可控制在怎样的距离下可以看到怎样级别的地形或影像。例如在地形中设置为3.0、2.9等常数，在影像中设置为TileDrawDistance和TileDrawSpread等常数。只是反映了一定的经验参数而已。
2 、 WorldWind 中对地块接边的处理是有裂缝的，采用裙摆设置的原理和方法实现是什么？
问题描述 ：对于不同级别的地形要进行接边是个很重要的问题，这个问题在我推荐的文献《全球多分辨率虚拟地形环境关键技术的研究》（作者杜莹2005博士）中有详细说明，而WW只是选择了最简单最原始的接地处理，也就是常说的裙摆处理。这种处理方法有很多的不利之处，但是实现方法简便，在此做简单说明。
问题切入 ：要想追踪这个问题，首先想到的是QuadTile的Update，而不从Render着手。Update就像在炒菜前把所有材料都准备的工作，如果没准备好就不能炒。在QuadTile的Update中，从CreateTileMesh切入，发现了CreateElevatedMesh和CreateFlatMesh的区别，显然裙摆是针对地形网格而做的处理在CreateElevatedMesh再深入到子函数，其中有一段很重要的说明：
/// <summary>
/// Builds flat or terrain mesh for current tile//为当前瓦片创建平坦或高程格网
/// </summary>
public virtual void CreateTileMesh()
{
verticalExaggeration = World.Settings.VerticalExaggeration;
m_CurrentOpacity = QuadTileSet.Opacity;
renderStruts = QuadTileSet.RenderStruts; if (QuadTileSet.TerrainMapped && Math.Abs(verticalExaggeration) > 1e-)
CreateElevatedMesh();
else
CreateFlatMesh();
}
在CreateElevatedMesh再深入到子函数，其中有一段很重要的说明：
/// <summary>
/// Build the elevated terrain mesh
/// </summary>
protected virtual void CreateElevatedMesh()
{
isDownloadingTerrain = true;
// Get height data with one extra sample around the tile
double degreePerSample = LatitudeSpan / vertexCountElevated;
TerrainTile tile = QuadTileSet.World.TerrainAccessor.GetElevationArray(North + degreePerSample, South - degreePerSample, West - degreePerSample, East + degreePerSample, vertexCountElevated + );
float[,] heightData = tile.ElevationData; int vertexCountElevatedPlus3 = vertexCountElevated / + ;
int totalVertexCount = vertexCountElevatedPlus3 * vertexCountElevatedPlus3;
northWestVertices = new CustomVertex.PositionNormalTextured[totalVertexCount];
southWestVertices = new CustomVertex.PositionNormalTextured[totalVertexCount];
northEastVertices = new CustomVertex.PositionNormalTextured[totalVertexCount];
southEastVertices = new CustomVertex.PositionNormalTextured[totalVertexCount];
double layerRadius = (double)QuadTileSet.LayerRadius; // Calculate mesh base radius (bottom vertices)
// Find minimum elevation to account for possible bathymetry
float minimumElevation = float.MaxValue;
float maximumElevation = float.MinValue;
foreach (float height in heightData)
{
if (height < minimumElevation)
minimumElevation = height;
if (height > maximumElevation)
maximumElevation = height;
}
minimumElevation *= verticalExaggeration;
maximumElevation *= verticalExaggeration; if (minimumElevation > maximumElevation)
{
// Compensate for negative vertical exaggeration
minimumElevation = maximumElevation;
maximumElevation = minimumElevation;
} double overlap = * verticalExaggeration; // 500m high tiles // Radius of mesh bottom grid
meshBaseRadius = layerRadius + minimumElevation - overlap; CreateElevatedMesh(ChildLocation.NorthWest, northWestVertices, meshBaseRadius, heightData);
CreateElevatedMesh(ChildLocation.SouthWest, southWestVertices, meshBaseRadius, heightData);
CreateElevatedMesh(ChildLocation.NorthEast, northEastVertices, meshBaseRadius, heightData);
CreateElevatedMesh(ChildLocation.SouthEast, southEastVertices, meshBaseRadius, heightData); BoundingBox = new BoundingBox((float)South, (float)North, (float)West, (float)East, (float)layerRadius, (float)layerRadius + * this.verticalExaggeration); QuadTileSet.IsDownloadingElevation = false; // Build common set of indexes for the 4 child meshes
int vertexCountElevatedPlus2 = vertexCountElevated / + ;
vertexIndexes = new short[ * vertexCountElevatedPlus2 * vertexCountElevatedPlus2 * ]; int elevated_idx = ;
for (int i = ; i < vertexCountElevatedPlus2; i++)
{
for (int j = ; j < vertexCountElevatedPlus2; j++)
{
vertexIndexes[elevated_idx++] = (short)(i * vertexCountElevatedPlus3 + j);
vertexIndexes[elevated_idx++] = (short)((i + ) * vertexCountElevatedPlus3 + j);
vertexIndexes[elevated_idx++] = (short)(i * vertexCountElevatedPlus3 + j + ); vertexIndexes[elevated_idx++] = (short)(i * vertexCountElevatedPlus3 + j + );
vertexIndexes[elevated_idx++] = (short)((i + ) * vertexCountElevatedPlus3 + j);
vertexIndexes[elevated_idx++] = (short)((i + ) * vertexCountElevatedPlus3 + j + );
}
} calculate_normals(ref northWestVertices, vertexIndexes);
calculate_normals(ref southWestVertices, vertexIndexes);
calculate_normals(ref northEastVertices, vertexIndexes);
calculate_normals(ref southEastVertices, vertexIndexes); isDownloadingTerrain = false;
}
这段注释的意思是说在：创建地形mesh的时候，在网格点的四周围多加一个点用于之后法向量的计算，而使用这种柱状顶点的效果将在法向量计算完后再折叠起来。这就是所谓的裙摆效果。在接下来的calculate_normals中，一眼就能看出折叠函数ProjectOnMeshBase，其具体实现如下：
// Project an elevated mesh point to the mesh base
private Point3d ProjectOnMeshBase(Point3d p)
{
p = p + this.localOrigin;
p = p.normalize();
p = p * meshBaseRadius - this.localOrigin;
return p;
}
声明：本文的原创著作权利属于武汉大学GIS(地理信息系统)专业2009级本科毕业生杨建顺学长，本文的成文仅用于个人学习和研究，在此对原著者表示感谢。
