Tinman 3D offers a unique set of features. Here are the key ones:
- Raster Datasets
Tinman 3D has its own highly optimized storage format for geodata rasters and giga-textures. Datasets support lossless compression, sparse data and dynamic updates. This takes away all groundwork when importing geodata; Tinman raster datasets will just grow and adapt as you import your data.
- Unlimited Detail
The maximum size of a geodata raster or texture that Tinman 3D can use is 2^30+1 by 2^30+1 samples. In other words, Tinman 3D will easily manage a terrain texture or DEM of 1.073.741.825 by 1.073.741.825 samples for you. With these dimensions, the Earth can be captured at a ground sample distance of 9mm, globally.
- Virtual Datasets
Datasets do not have to be provided as file-based rasters, data can also be computed at runtime. This is great if you have a procedural model that computes geodata or if you are using vector data to shape the terrain (for example roads).
- Runtime Data Processing
With Tinman 3D, there are almost endless possibilities on how to combine, merge and adapt geodata at runtime. Once you have your datasets ready, you can plug them together at runtime in just the way you need them to.
- Memory/CPU Consumption
You specify how much memory and CPU resources Tinman 3D shall use, and it will adhere to your specifications. The other way around, Tinman 3D will scale with your system and can utilize all memory and CPUs on a high-end multi-processor machine.
Tinman 3D uses rasters as input for creating terrain meshes and textures.
The basic process of creating a terrain mesh from raster data is simple:
The full resolution raster is stored using an efficient storage scheme.
The application defines a selection criterion, the terrain engine chooses raster samples accordingly. This creates the reduced raster.
The samples of the reduced raster are arranged into a graphics primitive (e.g. a single contiguous triangle strip). This yields a GPU-friendly triangulated irregular net (TIN).
Tinman 3D provides powerful implementations for these steps.
|1. Full Resolution Raster||2. Reduced Raster||3. Terrain Mesh|
The dynamic Continuous Level-Of-Detail (CLOD) mode of Tinman 3D is optimized for real-time applications.
The basic meaning of CLOD is that the whole terrain is represented by a single triangle mesh. The mesh triangles may differ in size, which makes it possible to capture different levels of detail at once.
A CLOD mesh is always contiguous, it does not have holes or T-junctions. Terrain meshes created by Tinman have an inherent quadtree structure, where each node represents a mesh sector. All terrain operations can be performed on the whole mesh or on specific sectors.
Using a CLOD approach for terrain rendering has various benefits:
- High Detail
CLOD can capture high local detail. If necessary, terrain mesh triangles can become very small - even down to a single pixel. This makes it possible to encode terrain data into mesh vertices which otherwise would have to be packed into textures. Effectively, this will save GPU memory and bandwidth.
- Low Detail
Only few triangles are used to represent terrain regions with low detail. The spared triangles will be used to capture high detail instead.
- Vertex Colors
The terrain surface texture can be modulated by using per-vertex diffuse colors, at any location, at any resolution. This makes dedicated modulation textures obsolete, which are limited in resolution anyway.
- Vertex Materials
Surface materials can be encoded as per-vertex weights. Usually material weights are used to texture near terrain parts. Materials are also used to control placement of ground objects (grass, plants, rocks, etc.). Tinman encodes independent weighted combinations of four materials out of a palette 256 possible materials, for each vertex. This makes terrain detail textures obsolete.
- Vertex Normals
A CLOD mesh accurately represents the terrain surface. Because of this, per-vertex normal, tangent and bitangent vectors are accurate enough for being used in shading and lighting computations. It is not necessary to use separate normal textures. Lightmaps can also be encoded as per-vertex data.
- Smooth Mesh
Tinman performs cubic interpolation at runtime in order to create smooth terrain surfaces, independently of the resolution of the input terrain data. This provides high-quality per-vertex data, which in turn allows an application to use advanced rendering techniques (for example, displacement mapping using GPU tesselation) with great success.
Tinman 3D can easily represent the whole surface of the Earth with a ground sample distance of ~9mm.
Furthermore, Tinman 3D Rasters are sparse, so an empty raster will not consume a single byte, independent of its size.
Current realtime 3D engines have an upper limit orders of magnitude below that. Here are some examples:
- Unreal Engine
Maximum "Recommended Landscape Size" : 8129 by 8129
Maximum "Recommended Terrain Size" : 4096 by 4096
Maximum "Heightmap Resolution" per Terrain object : 4097 by 4097
A typical game terrain spans an area of 16km by 16km. Using the maximum terrain sizes mentioned above, this would result in ground detail of 4m or 8m. Working around these limitations in order to improve ground detail requires considerable amount of manual work and design.
With Tinman 3D, you no longer have to worry about terrain size. You simply choose how big your terrain should be - then you decide how much detail you want to paint on it; on-the-fly and in real-time. The Demo Application uses a detail map of size 262144 by 262144. This would provide ground detail for a typical game terrain of 6cm. The possibility to paint arbitrary detail everywhere onto the terrain simplifies development and saves time.
The following images give an impression of what Tinman 3D can do.