Context: Procedural content generation shortened PCG is a way of letting the computer algorithmically generate data, with little input from programmers. Procedural content generation is a useful tool for developers to create game worlds, content and much more, which can be tedious and time-consuming to do by hand.Objectives: The procedural generation of both a city and height-mapped terrain parameter using Perlin noise and the terrain parameters effect on the city is explored in this thesis. The objective is to find out if a procedurally generated city with a heightmap parameter using Perlin noise is viable for use in games. Methods: An implementation generating both a height-mapped terrain parameter and city using Perlin noise has been created, along with that a user survey to test the generated city and terrain parameters viability in games. Results: This work successfully implemented an application that can generate cities affected with a heightmapped terrain parameter that is viable for use in games. Conclusions: This work concludes that it is possible to generate cities affected with a height-mapped terrain parameter by utilizing the noise algorithm Perlin noise. The generated cities and terrains are both viable and believable for use in games.

Machinists often simulate a part program to verify its correctness, since mistakes can cause damage to the part, machine, oneself, or others. A popular approach for part program simulation involves representing the stock (the material the part is being carved from) as a heightmap. Although this approach is computationally fast and memory efficient, only objects that are representable as functional surfaces (e.g., z = f(x, y)) can be machined. This thesis presents a new heightmap-based data structure, called a multidirectional heightmap, that does not have this limitation. A multidirectional heightmap, in response to an overhang, recursively subdivides itself until each piece can be represented by an axis-aligned heightmap. More precisely, a multidirectional heightmap is a kD-tree with the property that all cells are functional: each cell contains a heightmap that represents a functional portion of the stock. To improve accuracy, each regular heightmap can be replaced by a 3-Way Heightmap, a new type of heightmap that samples the tool along all three stock axis directions (three ways) rather than just one. The experimental results herein suggest that the multidirectional heightmap data structure achieves a good level of performance with respect to memory usage, CPU usage, and approximation error.

Machinists often simulate a part program to verify its correctness, since mistakes can cause damage to the part, machine, oneself, or others. A popular approach for part program simulation involves representing the stock (the material the part is being carved from) as a heightmap. Although this approach is computationally fast and memory efficient, only objects that are representable as functional surfaces (e.g., z = f(x, y)) can be machined. This thesis presents a new heightmap-based data structure, called a multidirectional heightmap, that does not have this limitation. A multidirectional heightmap, in response to an overhang, recursively subdivides itself until each piece can be represented by an axis-aligned heightmap. More precisely, a multidirectional heightmap is a kD-tree with the property that all cells are functional: each cell contains a heightmap that represents a functional portion of the stock. To improve accuracy, each regular heightmap can be replaced by a 3-Way Heightmap, a new type of heightmap that samples the tool along all three stock axis directions (three ways) rather than just one. The experimental results herein suggest that the multidirectional heightmap data structure achieves a good level of performance with respect to memory usage, CPU usage, and approximation error.

<p> </p><p>Implementering och analysering av existerande tekniker inom generering av höjdkartor (eng. Heightmap). Försök till att komplettera brister i dessa med en egen teknik. Samt jämföra dessa tekniker med verklig data från områden i USA.</p><p> </p>

Computer-aided generation of virtual worlds is vital in modern content production. To manuallycreate all the details which todays computers can visualize would be too daunting atask for any number of artists. Procedural algorithms can quickly generate content, but thecontent suffers from being repetitive. Simulation of geological processes produce good resultsbut require a lot of resources. In this report a solution is presented which combines procedural algorithms with geologicalsimulation in the form of erosion. A pre-processing stage generates a heightfield using proceduralnoise which is then eroded. The erosion is accelerated by being performed on the GPU.A road network is generated by connecting points scattered in the world. The pre-processedworld is then used to define a field function. The function is sampled in a grid as neededto produce voxels with different materials. Roads are added to the world by changing thematerial of the voxels. The voxels are then rendered as textured tiles depending on material. The generated worlds are varied and interesting, much more so than worlds created purelyby procedural methods. A world can be pre-processed within a few minutes and explored inrealtime.

Interactive software applications which need to render large terrain meshes can suffer from slow frame rates if the geometry of the terrain is sufficiently dense. However, the viewing angle to many distant features of the terrain does not change rapidly with respect to time. If the movement of the viewing position is limited to continuous motion and restrained to a known speed, many terrain features may be rendered once in high detail and reused for several frames. This thesis proposes a method to increase the rendering speed of large complex terrains by splitting the terrain into contiguous chunks. If a given chunk is far enough away from the camera and its viewing angle will not change quickly, it is rendered into an image buffer. This buffer is then used to texture map a simplified version of the terrain mesh. The simplified and textured mesh is rendered in place of the original chunk of geometrically complex terrain. The simplified mesh is used to approximate parallax effects as the viewing angle changes in small increments. This technique is shown to as much as double the rendering speed of large terrain meshes without reducing the quality of the final image.

Implementering och analysering av existerande tekniker inom generering av höjdkartor (eng. Heightmap). Försök till att komplettera brister i dessa med en egen teknik. Samt jämföra dessa tekniker med verklig data från områden i USA.

The subject matter of this work is a concept of an interactive 3D terrain editor. Its basic functionality, graphical user interface and possible extensions are defined. There are a lot of existing terrain editing tools, therefore presentation of some existing systems capable of editing and generating computer terrain is also included. The editor is implemented and it is capable of performing simple editing, such as raising, lowering or levelling the terrain. Graphical user interface is also a part of the implementation. The editor allows inserting 3D models into the scene so it was easier to represent some objects displayed only in the texture in 3D space. Models can be edited using some simple transformations. Performance tests are undertaken and possibilities of extending the program discussed.

The goal of this thesis is to design a suitable method for lossy compression of heightmap terrain data. This method should accept blocks of float samples of dimensions 2^n x 2^n as an input, for which it should be able to perform progressive decompression of mip-maps (lower-resolution representations). It should keep the reconstructed data within a certain maximum per-sample error bound for each mip-map level. This bound should be in the unit of meters and adjustable by the user. Given these constraints, it should be as efficient as possible. Our method is inspired by the second generation of progressive wavelet-based compression scheme modified to satisfy the~maximum-error constraint. We simplified this scheme by factoring out unnecessary computations in order to improve the efficiency. Our method can compress a 256x256 block in about 30 ms and decompress it in about 2 ms. Thanks to these attributes, the method can be used in a real-time planet renderer. It achieves the compression ratio of 37:1 on the whole Earth 90m/sample terrain dataset transformed and separated into square blocks, while respecting the maximum error of 5m. Powered by TCPDF (www.tcpdf.org)

Deep learning has proven a powerful tool for image analysis during the past two decades. With the rise of high resolution overhead imagery, an opportunity for automatic geospatial 3D-recreation has presented itself. This master thesis researches the possibil- ity of 3D-recreation through deep learning based image analysis of overhead imagery. The goal is a model capable of making predictions for three different tasks: heightmaps, bound- ary proximity heatmaps and semantic segmentations. A new neural network is designed with the novel feature of supplying the predictions from one task to another with the goal of improving performance. A number of strategies to ensure the model generalizes to un- seen data are employed. The model is trained using satellite and aerial imagery from a variety of cities on the planet. The model is meticulously evaluated by using four common performance metrics. For datasets with no ground truth data, the results were assessed visually. This thesis concludes that it is possible to create a deep learning network capa- ble of making predictions for the three tasks with varying success, performing best for heightmaps and worst for semantic segmentation. It was observed that supplying estima- tions from one task to another can both improve and decrease performance. Analysis into what features in an image is important for the three tasks was clear in some images, unclear in others. Lastly, validation proved that a number of random transformations during the training process helped the model generalize to unseen data. / <p>Examensarbetet är utfört vid Institutionen för teknik och naturvetenskap (ITN) vid Tekniska fakulteten, Linköpings universitet</p>