Windfarm Occlusion Python Application

This Python project originates from a bachelor thesis: Occlusion for Augmented Reality Windfarms in Great Distance from January 2023. It is one of two code projects of a developed prototype to evaluate different computer vision algorithms along topographical data to solve the occlusion problem for wind farm visualization at great distances. It was developed to evaluate a Dataset which was created using the Unity prototype.

Requirements

• DenseDepth git repository: git clone https://github.com/ialhashim/DenseDepth.git
• DenseDepth Neural Network: !wget https://s3-eu-west-1.amazonaws.com/densedepth/nyu.h5 -O ./DenseDepth/nyu.h5

Conda environment:

$ conda create -n depthEnv jupyter tensorflow-gpu matplotlib pillow pandas keras scikit-learn scikit-image pydot tqdm seaborn

$ conda activate depthEnv

$ conda install -c conda-forge opencv

Folder Sturcture:

• Dataset: Contains the Visualization of the dataset and its occlusion results.

• FullPaulusstr: Contains the Scenes that were captured from the location Paulusstr that were used to compute the visualizaiton of the Dataset folder.

• Informatikzentrum: Contains the Scenes that were captured from the location Informatikzentrum that were used to compute the visualizaiton of the Dataset folder.

• Koenigshuegel: Contains the Scenes that were captured from the location Königshügel that were used to compute the visualizaiton of the Dataset folder.

• Lousberg: Contains the Scenes that were captured from the location Lousberg that were used to compute the visualizaiton of the Dataset folder.

• Handler: Contains all python files and classes.

• Presentation: Contains some jupyter notebooks that were used to create visualizations for the bachelor defense.

Project Structure

• OcclusionHandler: Central point of this project. Uses the FileHandler, SfMHandler, SkyHandler and MDEHandler to provide all needed occlusion maps and also to provide the occlusions.

  • FileHandler: Loads all needed files and adjusts them accordingly and providing the OcclusionHandler access to those files.
    • RotationHandler: Used by the FileHandler to rotate and cut the images as needed.
  • SfMHandler: Provides OcclusionHandler with the SfM occlusion map.
  • SkyHandler: Provides OcclusionHandler with the Sky occlusion map.
  • MDEHandler: Provides OcclusionHandler with the MDE occlusion map.

• utils: Provides different utility functions.

• AnalysisHandler: Is a wrapper around the OcclusionHandler, and can be used to compute different metrics of a Scene.

• SequenceAnalysisHandler: Is a wrapper around the OcclusionHandler, and uses the AnalysisHandler to compute different of all Scenes for an entire Sequence.

Standard Procedure

OcclusionHandler is initialized for a specified Sequence. With load_scene a secified Scene is loaded. During that the FileHandler is used to load all data from the specified Scene and adjusting it, using the RotationHandler to also provide the rotated and cutout images. After that the OcclusionHandler uses the data provided by the FileHandler to initialize the SfMHandler, the SkyHandler and the MDEHandler. Now the different occlusion maps can be accessed using the OcclusionHandler.