Humanities & Industry
Large numbers of historical handwritten texts have now been digitised. We develop methods to analyse and explore these images. Three dimensional analysis of wood fibre based materials, especially composites, is one of our specialities. We also develop new better ways to display the railway network to traffic controllers and to help industry use image analysis for quality control.
Even though a major fraction of vi2's projects fall within health care, medicine and biology, we also have projects in a wide range of other application fields, most of which we can classify as either humanities or industry.
Hand-written text recognition
In collaboration with the Uppsala university library, the faculty of arts, the faculty of languages and the IT department, we explore automatic recognition in historical hand-written manuscripts. Already, large collections of hand-written material is available as digital images. However, there is gap in the technology for searching and translating these into digital text. We are working on techniques to enable searching and data mining in large image databases with historical hand-written documents. Our methods will enable both researchers and the public to access historical sources in a similar way as we access other large collections of digital text on the internet today. This work has the potential to revolutionize the humanities. As of today, manual transcription of these hand-written manuscripts can only proceed at the pace of 3-5 pages per day.
Together with researchers at Stockholm university, we develop methods to extract land use parameters and urban areas from satellite images. This help researchers in geography to use high-resolution satellite images to quantify their research. In collaboration with IIT, CNR in Pisa, Italy both new and improved methods are developed for efficient correction and stitching of digitized historical aerial photos in the GeoMemories project, which aims at creating a web application covering the spatial-temporal evolution of the Italian landscape. The image to the left shows how three aerial photos over Pisa have been stitched together and one can see the leaning tower under the cathedral. The photos are from two different flights by RAF over Pisa during WWII. (C) MiBAC-ICCD, Aerofototeca Nazionale, fondo RAF.
At Vi2 we have a strong connection to industry. Traditionally, image analysis has been used to control production lines in factories, by directing the robots or by examining the output of the line. Furthermore, information about complex production processes needs to be presented to human operators in ways that allow them to oversee the process and quickly react to changes. Many of our projects in connection with industry are in the form of master theses or consultancy, but we also have some larger projects.
For example, we have been working on microstructural analysis of wood-fibre-based materials since the late 1990s. Forestry is one of the largest industries in Sweden, and a natural way of increasing the value of its output is to develop new, high-tech materials based on wood. We are using micro-computed tomography and electron microscopy to study the three-dimensional structure of wood-fibre composites at different scales, in order to learn how to tailor the production of these materials for different applications. Image to the right: Rendering of the fibres in a wood fibre-plastic composite, imaged using synchrotron radiation micro-CT. The block is one cubic micrometer.
More traditional uses of image analysis in industry are represented by a project with Vattenfall, in which we monitor particle aggregation in fluidized bed boilers, or by a project with Lantmännen Lantbruk, in which we develop a vision-based machine to monitor the quality of grain harvests. A large project exemplifying human monitoring and control of complex processes is our Future Train Traffic Control project. Here, we are developing new ways to display information about all trains in the railway network, and allowing operators to modify the train schedule in real time. Image to the left: An operator using STEG to make sure that trains are running according to plan.