Warning
An improved implementation of this code can be found here.
This repository is related to the RegBL (RBD/GWR/RegBL) completion research project. The STDL was contacted by the Swiss Federal Statistical Office (OFS) to determine in which extend it could be possible to complete the construction date of Swiss buildings based on the analysis of a temporal sequence of the Swiss federal maps produced by swisstopo. With an initial target of 80% of correct guesses, the goal of this research project was to demonstrate the possibility to reach such goal using a reliable validation metric.
This repository holds the primary pipeline codes used to process the RegBL entries using the swisstopo maps. It also comes with a specific tool used to create temporal representation of the situation of buildings using maps and RegBL information.
The following links give access to the codes related to the project :
- Primary pipeline - Construction dates extraction using maps (This repository)
- Secondary pipeline - Construction dates extraction without maps
- Results and analysis tools for the primary pipeline
The following links give access to official documentations on the considered data :
In the first place, the primary pipeline is split into step-processes that are each responsible of achieving a specific part of the overall process. The following links gives access to the documentation of each of the step-processes :
- Source map segmentation
- Processing structure bootstrap
- Building detection
- Construction date deduction
- Timelines composer
The pipeline works by considering a main storage directory in which everything is stored, including source maps, segmented maps, intermediate results and final results. This directory has to be created in the first place :
$ mkdir -p .../regbl_process
The geographical area on which the process has to take place needs to be defined. It has to be a geographical region on which historical maps are available. For example, the Bern area used during the research project is defined in terms of its boundaries (EPSG:2056), giving the easting (longitude) and northing (latitude) minimum and maximum values :
2596000 2602000 1197000 1203000
A sequence of historical maps, available in geotiff, covering the selected area has to be gathered. In the case of this research project, we used the swisstopo 1:25'000 national maps (KOMB). We highly recommend to use maps with the same scale and coming with a geographical ratio identical to the ratio of the geographical area (1:1 in terms of easting and northing pixel geographical edges). In addition, the segmentation process was designed to work on the swisstopo maps. Some adaptations could be required to work on other maps.
The source maps have to be cropped according to the defined geographical area. We recommend using the GDAL tool to perform the geographical crop. For the Bern example, one can use :
$ gdalwarp -of gtiff -te 2596000 1197000 2602000 1203000 source_map.tif [year].tif
The original cropped maps have to be placed in the following directory (the TIFF format is mendatory) :
.../regbl_process/regbl_frame/frame_original/[year].tif (2010.tif, 2004.tif ...)
As an illustration, the following image gives a sample of the cropped maps chosen to cover the Bern area during this research project. As one can see, the maps are square, as the selected geographical area (1:1 pixel edges) :
Sample of the considered maps of Bern (1993, 1998, 2004, 2010) - swisstopo, 1:25'000 national maps
The next step is to apply the segmentation process on each of the selected map to clean them and to extract the pixel footprint of the buildings. The maps have to be processed one by one using the segmentation tool :
$ ./regbl-segmentation -i .../frame_original/[year].tif -o .../frame/[year].tif -e
and stored in the :
.../regbl_process/regbl_frame/frame/[year].tif (2010.tif, 2004.tif ...)
directory and named after their respective year. These segmented maps are the one used to performed the detection and deduction of the building construction date. The following image gives and illustration of the obtained segmented (pre-processed) maps :
Result of the segmentation step-process applied on the previous maps
A last element is needed to fully initialise the main storage directory, that is the 3D raster descriptor file. It is a simple text file that gives, line by line, the year of the maps, its geographical boundaries and the width and height (in pixels) of the corresponding segmented map. For example, the Bern case considered in this research project has the following descriptor :
2010 2596000 2602000 1197000 1203000 4800 4800
2004 2596000 2602000 1197000 1203000 4800 4800
1998 2596000 2602000 1197000 1203000 4800 4800
1993 2596000 2602000 1197000 1203000 4800 4800
...
The entries have to be given in the decreasing temporal order. The file storing these information as to be named regbl_list and place in the main storage directory.
As the main directory is prepared as described, the three successive processing steps can take place, starting with the bootstrap one :
$ ./regbl-bootstrap -s .../regbl_process -g .../RegBL/GEB.dsv [-e .../RegBL/EIN.dsv]
which reads the RegBL GEB and EIN databases to extract the relevant buildings information based on the defined geographical area through the 3D raster descriptor. Note that specifying the EIN database is not mandatory. In case the EIN database is not provided, the entries of the buildings are not considered for the deduction of their construction date.
The detection process can then takes place :
$ ./regbl-detect -s .../regbl_process
which performs the building detection on each segmented map (3D raster slices). The deduction process can then takes place :
$ ./regbl-deduce -s .../regbl_process
which performs the final deduction of the building construction temporal range. The results of this last step can be found in the :
.../regbl_process/regbl_output/output_deduce
where a large amount of files, one per building, can be accessed to read the upper and lower boundaries of the deduced construction range. The files are all named after the EGID of their building.
This repository also offers a way to represent the situation of a building across the 3D raster along with information on the detection and deduction processes. Based on a given EGID, the tracker can be used, on a fully processed main storage directory, to compute the representation :
$ ./regbl-tracker -s .../regbl_process -g 1269938 -e .../export/directory
The representation is a PNG image named after the building EGID. The following image gives an example of such representation :
Example of building situation and results representation
The scripts proposed by the results and analysis tool can also be considered to analysis the results more in details.
regbl-poc - Nils Hamel, Huriel Reichel
Copyright (c) 2020-2021 Republic and Canton of Geneva
This program is licensed under the terms of the GNU GPLv3. Documentation and illustrations are licensed under the terms of the CC BY 4.0.
The regbl-poc comes with the following package (Ubuntu 20.04 LTS) dependencies (Instructions) :
- build-essential
- libopencv-dev
The code documentation is built using Doxygen.
To build the project, including the sub-modules, use make with the following targets :
$ make clean-all && make all
To rebuild the binaries without rebuilding sub-modules, use the make targets :
$ make clean && make build
To only rebuild sub-modules, use the make command with the targets :
$ make clean-module && make module
To generate the documentation of the project, use the targets :
$ make clean-doc && make doc
and the targets :
$ make clean-all-doc && make all-doc
to generate the documentation of the project and its sub-modules.