Information Systems, AI and Circular Economy
Topics of the track
Governmental Information Systems Supporting Nature Conservation in Schleswig-Holstein
The Ministry of Energy, Agriculture, the Environment, Nature and Digitalization Schleswig-Holstein (MELUND) is responsible for the management of nature conservation information in Schleswig-Holstein supported by the State Agency for Agriculture, the Environment and Rural Areas (LLUR). Both authorities need reliable and up-to-date data in order to fulfil their governmental tasks.
To develop and operate comprehensive information systems supporting – amongst others - the management of nature conservation data the governmental strategy recommends the hosting and service environment of the highly secure central data centre. This paper describes design and architecture of a topical cluster for nature conservation data within the central data centre.
Landfill site selection using GIS and Remote Sensing
Finding and selecting the landfill site for disposing of the solid waste is a serious challenge for a metropolitan like Kathmandu. Due to improper management of waste materials, different health-hazardous diseases are growing. Global warming and methane gas production are also a serious causes of poor disposal which are badly affecting the environment and ecology. There are different methods for selecting the landfill site. The policy 3R viz. reduction, recycle, and reuse adopted by metropolitan should be considered while disposing of the wastes. The use of GIS, Remote sensing and analytic hierarchy become crucial while selecting landfill sites. We obtain the satellite imagery covering
Kathmandu and analyzed using GIS to determine geologically and geographically suitable places. GIS performs some deterministic overlay and buffer operations. Almost 12 criteria were used. Distance from waste generation center, distance from roads, slope, distance from settlements, distance to surface water, distance to groundwater areas, soil permeability, etc. are some of the criteria used. These criteria are given some relative weight according to their importance using analytic hierarchy. The map of a suitable site is prepared using GIS spatial operations, rank the candidate site and the most suitable place is recommended.
A Blueprint for Computer Vision Testing in the Circular Economy
Automating visual inspections of used parts constitutes a challenge impeding wider real-world implementations of ideas from the circular economy (CE). Computer vision (CV) is a promising technology in this regard, but its real-world testing is a technical and economic challenge. This CV testing blueprint is an experimental and practice-proven approach for evaluating whether CV is a technically and economically suitable tool for automating a visual inspection task. The testing of two hypotheses lies at the core of the blueprint: (A) The task is solvable based on digital image data and this data is obtainable, and (B) available CV algorithms can learn the task. While testing Hypothesis A does not require machine learning know-how, testing Hypothesis B uses a new technique to infer the prospective algorithm performance from the shape of the validation accuracy graph when trained on a subset of the required training data.
Garment attribute prediction using camera images and Raman spectroscopy to enhance circular textile economy.
In the last decade, demand in the fast fashion segment for textile fibers has increased remarkably. The global fiber consumption is expected to reach between 130 and 145 million metric tonnes per year by 2025.
This forecast comes with a huge negative impact on the environment, and it is clear that to minimize the side effects of this industry, a major transformation must occur.
One of the main challenges is to transform the industry from its primarily linear to a circular structure so that it can reuse its raw materials rather than disposing them at the end of use. Closing the loop would mean a significant reduction of greenhouse gas emissions. Recycling garments and turning them into raw material is still a challenging task; among other difficulties, it requires exact information about the composition of the garments to be recycled.
Currently, conventional sorting plants focus on extracting reusable garments to be sold to the second-hand market while paying for the disposal of the non-reusable fraction. However, this task is complex; sorting an enormous amount of mixed bales of worn clothing with up to hundreds of different attributes is nowadays done by hand.
The presented project aims to reliably and objectively sort single garments according to a set of defined attributes, detect the material composition and contaminants, and enable fiber-to-fiber recycling. In this way, clothing contaminated with harmful substances could be sorted out and sent for hazardous waste recycling, thus preventing contamination in the recycling process or contamination to the environment due to incorrect hazardous disposal.
Chair: Klaus Greve