Provide optimized and robust methods to detect and decode aztec codes by using opencv and zxing-cpp in combination and to
transcode UIC918 information with signature validation into json structure. (UIC918-3 and UIC918-9)
Looking for build instructions? Take a look at the end of this document!
Decoder can detect and decode uic918 content from aztec codes from given input images/pdfs, verifies content and prints json result on stdout or dumps it into file.
Check ticket-decoder --help for arguments.
Provided python API is in an early state and the class DecoderFacade supports 2 methods right now only.
decode_uic918('...')is considered for the use case you decode the raw data from aztec-code in advance via zxing or other aztec-code-decoder of your choice and you want to decode raw UIC918 data to json only. In this case, be careful with the output of the decoder to avoid string encodings like UTF8 or other multi-byte encodings. Ideally try to get access to the raw byte array and just encode those bytes to base64 before passing it to the method. If your aztec-code-decoder provides a string-type only and you are able to pass character-encoding, try using 'ISO 8859-1' and cast the result string to raw bytes.decode_files('...')detects and decodes aztec-codes from file or files (pdf, image) and decodes raw UIC918 data to json. This is using zxing-cpp internally. It returns an array of tuples (input-path and json-result) of size x, while x is the amount of aztec-codes found on input.
To build the module, some tools and dependencies are required. Beside python3 and essential build tools, it is required to have python3-dev installed. In Ubuntu, the following steps should be enough to get it built.
apt-get install --no-install-recommends -y build-essential cmake python3-pip python3-dev python-is-python3
pip3 install "conan<2.0" "numpy<2.0"
conan profile new ticket-decoder --force --detect
conan profile update settings.compiler.libcxx=libstdc++11 ticket-decoder
. ./setup.Python.sh
Ensure PYTHONPATH is defined to enable Python to discover the ticket_decoder module. Try executing the test-cases.
export PYTHONPATH=`pwd`/build/Release/bin
python3 -m unittest discover -s source/test/python/
When the module has been build successfully, a Python script as shown below should work.
See source/python/run.py or source/test/python/test_decode_uic918.py for more detailed examples.
from ticket_decoder import DecoderFacade
decoder_facade = DecoderFacade()
for result in decoder_facade.decode_files('path/2/your/ticket.pdf'):
print(result[0] + ": " + result[1])
Analyzer is able to scan for aztec codes in images grabbed from camera or from images/pdfs in a folder. It provides a simple interactive mode to visualize detection, image processing and decoding steps and to change some parameters to find optimal setup for detection. This application is considered to optimize default parameters and algorithms for the decoder.
Check ticket-analyzer --help for arguments.
To get a minimal setup for experimentation, do the following:
- Download UIC918-3 and UIC918-9 sample tickets from https://www.bahn.de/angebot/regio/barcode and extract zip files into folder ./images/
- Download XML file containing public keys of issuers from https://railpublickey.uic.org/ into folder ./cert/ and name it UIC_PublicKeys.xml
- Run
./build/Release/bin/ticket-analyzerfrom workspace folder or use arguments to specify different paths to input files and folders - Use following keys to tweak settings:
- i: Visualize next image processing step
- I: Visualize previous image processing step
- c: Visualize next contour detection step
- C: Visualize previous contour detection step
- f: Next image input file from image-folder
- F: Previous image input file from image-folder
- : Toggle camera device (space)
- r: Rotate image -1 degree
- R: Rotate image +1 degree
- 2: Split image into 2 parts and rotate over parts
- 4: Split image into 4 parts and rotate over parts
- s: Scale up image
- S: Scale down image
- 0: Reset: Rotation, Scale, Split
- d: Rotate over available detector implementations
- p: Assume pure barcode
- b: Use local average binarizer
- D: Dump current image into output-folder
- o: Overlay detected barcode image
- t: Overlay decoded content or text
- Check output-folder (./out by default) for intermediate images, raw data files or decoded data in json files
- Printed code size: 48mm (1.89inch)
- With 200dpi: 1.89 inch/code * 200 dot/inch ~ 380 dot/code
- With UIC-918-3: 380 dot / 87 blocks ~ 4.37 dot/block
- Recommendation on TAP TSI Revision - Technical Document - B12
https://www.era.europa.eu/system/files/2022-10/Recommendation%20on%20TAP%20TSI%20Revision%20-%20Technical%20Document%20-%20B12.pdf
(Link not working anymore? Go to https://www.era.europa.eu/ and search for "TAP TSI Revision - Technical Document - B12")
(TAP: Telematics Applications for Passenger Service)
(TSI: Technical Specifications for Interoperability)
- UIC-barcode https://github.com/UnionInternationalCheminsdeFer/UIC-barcode (Apache License 2.0)
- Install free open source ANS.1 compiler (BSD 2)
https://github.com/vlm/asn1c
- MacOS:
brew install asn1c - Ubuntu:
apt install -y asn1c
- MacOS:
- Generate required code by using the following support script:
# Clones the repository and calls asn1c with matching parameters at the right places
#
./etc/setup.uic-asn1.sh
- Interoperabilität Barcode DB Online-Ticket
https://assets.static-bahn.de/dam/jcr:8fa0c0b5-d7b8-443b-b3cd-7ae902884847/236539-315207.pdf
- Parser für Onlinetickets der Deutschen Bahn
https://github.com/rumpeltux/onlineticket (GPL 3.0) - uic-918-3
https://github.com/justusjonas74/uic-918-3 (MIT)
-
Public keys from UIC
https://railpublickey.uic.org/ -
List of numeric codes for railway companies (RICS Code)
https://uic.org/support-activities/it/rics -
DB Railway Station Documentation (EVA-Nummern)
https://data.deutschebahn.com/dataset/data-haltestellen.html
-
Interoperability UIC/VDV codes, UIC918-3 and UIC918-9 example tickets and mappings for ids used in VDV codes
https://www.bahn.de/angebot/regio/barcode# You can use the following command to convert PDF file into images for further processing, but you don't have to because application is able to precess pdf files directly. But decoding quality might differ depending on parameters like DPI. # brew|apt install imagemagick convert -density 250 -trim -quality 100 -flatten <file name>.pdf <file name>.png -
DB-AGs OLT Barcode to VDV Data Structure Reference Implementation
https://sourceforge.net/projects/dbuic2vdvbc/ -
HandyTicket-Fahrausweise des VRR im VDV-Barcode
https://www.kcd-nrw.de/fileadmin/03_KC_Seiten/KCD/Downloads/Technische_Dokumente/Archiv/2010_02_12_kompendiumvrrfa2dvdv_1_4.pdf -
Additive Datenübertragung in Barcodes von internationalen Bahntickets
https://monami.hs-mittweida.de/frontdoor/deliver/index/docId/4983/file/WaitzRoman_Diplomarbeit.pdf -
KDE Barcode Formats - Ticket Barcode Formats
https://community.kde.org/KDE_PIM/KItinerary/Barcode_Formats Some details and collection of links related to different european rail companies and their ticket formats.
-
gcc >= 11, clang >= 14 (other compilers and versions may work but are not tested)
-
conan package manager < 2.0 (https://conan.io/)
-
cmake >= 3.19
-
python3 numpy (boost.python requires numpy for build and unfortunately, it is not possible to disable it via conan config)
Following libraries are used by the project. Usually you should not care about it since conan will do that for you.
- opencv (image processing)
- zxing-cpp (barcode/aztec-code decoding)
- nlohmann_json (json support - output)
- easyloggingpp (logging)
- pugixml (xml support - public key file)
- botan (signature verification)
- tclap (cli argument processing)
- gtest (unit testing)
- poppler (pdf reading/rendering)
- boost.python (python binding)
In general, when you want to avoid to install additional dependencies like non-default compilers and libraries on your system, consider using one of the build scripts using a docker container to create the build environment.
As long as the conanfile.py is unchanged, you can re-use the container with pre-built dependencies, source code changes are directly mirrored into build environment and artifacts are mirrored back into host system. In case dependencies change, the container gets re-build with updated dependencies.
-> this will install dependencies and run the build inside a ubuntu docker container
- setup.docker.ubuntu22.gcc11.sh
- setup.docker.ubuntu22.clang15.sh
- setup.docker.ubuntu22.gcc11.Python.sh
When the preparation of the build environment has been successful, it should be possible to build the project by using ./build.sh -j inside the build container.
Take a look into ./build/ folder to discover artifacts. You should be able to execute the executables on host machine as well.
When opencv has to be built from source because of missing pre-built package for your arch/os/compiler mix, it might be necessary to install some further xorg/system libraries to make highgui stuff building inside conan install process. To get this handled properly, use the following conan config flags:
- conf.tools.system.package_manager:mode=install
- conf.tools.system.package_manager:sudo_askpass=True
as shown below OR install ALL required xorg dependencies manually. For details about specific required packages please check the error message carefully or see the step "Install compiler and stdlib" in ".github/workflows/c-cpp.yml" for a list of dev-package names.
apt-get install --no-install-recommends -y build-essential make cmake git wget python-is-python3 python3-pip python3-dev libgtk2.0-dev
pip3 install "conan<2.0" "numpy<2.0"
conan profile new --detect --force ticket-decoder
conan profile update settings.compiler.libcxx=libstdc++11 ticket-decoder
conan profile update conf.tools.system.package_manager:mode=install ticket-decoder
conan profile update conf.tools.system.package_manager:sudo_askpass=True ticket-decoder
git clone https://github.com/karlheinzkurt/ticket-decoder.git
cd ticket-decoder
./setup.Release.sh -- -j
etc/install-uic-keys.sh
build/Release/bin/ticket-decoder-test
etc/python-test.sh
It might be required for dependencies to get built properly during conan install to have a
python command (without 3) in path available. So when you face an error like python: command not found
it might be required to create a link via sudo ln -s $(which python3) /usr/local/bin/python since there
is no package python-is-python3 in homebrew available, as it is for ubuntu.
xcode-select --install
brew install cmake
pip3 install "conan<2.0" "numpy<2.0"
conan profile new --detect --force ticket-decoder
conan profile update settings.compiler.version=15.0 ticket-decoder
git clone https://github.com/karlheinzkurt/ticket-decoder.git
cd ticket-decoder
./setup.Release.sh -- -j
etc/install-uic-keys.sh
build/Release/bin/ticket-decoder-test
etc/python-test.sh
For sure, it should be possible to get it built by using visual compiler and toolchain as well. But I never tried and you might need to modify some build parameters/arguments and you have know (or to find out) how to setup toolchain, conan and cmake in Windows environment. Furthermore, the compiler might complain about things gcc and clang are not complaining about. But when you are an experienced dev, you should be able to get it managed. (support of multiple u_flex versions via asn1c generated and unprefixed C source files in a shared lib makes this a bit harder, most probably, since export/import of shared libs has to be ported to visual compiler world to, but it's possible via crazy macro stuff, i know)