Denne idé er en del af Banedanmark - Digitalisering af ledningstegninger
Banedanmark has an immense database of information going beyond 100 years of blueprints with cables. The digitalisation of that database is a tedious, resource-consuming process that is perceived as needed and, at the same time, as highly inefficient, due to the variety of formats that the digital blueprints database contains and the complexity of each image.
OMNIWIRE aims to solve that issue, offering a value proposition that entails innovative technology, efficiency and optimisation of resources through the implementation of a conscious digitalisation tool for the database, generating a potential unification of the database in a single extension file type. In that sense, OMNIWIRE represents a second-generation digitalisation tool, understanding that the transition from simple digitalisation (from paper to a digital file) to smart automatised digitalisation (from digital file too big data analytics) needs a “conscious” device with the affordance of understanding processed information and transform it into unified knowledge.
The program works as a conscious digitalisation tool which features an enhancement engine for optimisation of image quality. That optimisation is the first step, which will lead to a processing of 75% on the documents (PDF and TIFF formats). The other 25% (DWG and DGN) are scalable and, therefore, suitable to be enhanced without a need for enhancing tool.
As a second step, OMNIWIRE will apply computer vision for detecting the blueprints’ relevant information, for instance; wires, signs, codes and numbers, while neglecting the information not suitable to be processed, such as buildings or roads. Once detected, the relevant information is extracted and fed as input into a machine learning medium, which can optimise differential parameters across the different images.
In addition, a geo-location tool is added to the program. By establishing a known reference point, the orientation of the blueprint (geographical North) and scale of the model, OMNIWIRE can register the geolocation of the different wires and junctions by referencing to that initial established point (known reference point).
As the last step, the information is coded to facilitate that the machine learning medium can create the output data into a GIS (Geographical Information System) program.
When interacting with OMNIWIRE, the user is presented with an interface (image attached to this file) that demands an initial input as a way for identifying codes for wiring for the first time, which will then be added to a library of signs. By making use of the feature ‘SCAN’, OMNIWIRE can scan for similar library signs and codes to be found in the specific blueprint. As the program makes use of ML it will become more and more precise with every interaction in identifying the relevant information of the blueprints. For this purpose, the user needs to supervise the first version of blueprints by dragging symbolics and “lines” of wires onto the digital overlay.
By: Victor M. Angulo & Martin V. Højgrav-Huus