Software & Systems

Articles of journal № 3 at 2019 year.

Order result by: Public date | Title | Authors

21. Reconstruction of urban space texture model based on topographic maps and camera records [№3 за 2019 год]
Authors: A.P. Kudryashov, I.V. Solovev
Visitors: 4713
The tasks of reconstructing urban space scenes can use various materials as a data source: satellite im-ages, video series, optical system data, etc. The paper proposes an approach for reconstructing a three-dimensional model of urban space using the method of recognizing service information on a topographic map. Topographic maps are the basic data at all stages of architectural, planning and engineering design. They contain information on the building footing geometry and their position among other objects. For recognition, the authors use a modified wave algorithm, which allows identifying and recognizing closed contours in an image that are classified into various objects: building contours, labels, service symbols, etc. The paper presents rationale of advantages of the considered algorithm to select contours. The paper proposes a method of applying textures to three-dimensional models of buildings. The textures are from photographs of real buildings. It is proposed to use special textures for certain types of buildings in a case when there is no real photograph of the building. Photos are attached to a topo-graphic map using geographic coordinates. The authors also describe a method of binding reconstructed objects to a relief. There is an infor-mation system used both for the entire reconstruction process and to solve individual local tasks. The paper gives the examples of reconstructing real topographic maps of 1:2000 scale.

22. A simplified method for skeletonization of non-convex figures [№3 за 2019 год]
Author: A.V. Kuchuganov
Visitors: 3961
The approximation of graphic information through the skeletonization of object images is a way to re-place objects with simpler and more convenient representations in semantic analysis problems and im-age recognition. Skeletons are widely used in technical vision systems, content image search, in geo-metric modeling and visualization. The most popular approaches: based on “erosion” (removal of ob-ject boundary points) and mathematical (based on Voronoi diagrams formed by Delaunay triangula-tion, inscribing circles or using the wave method). A common disadvantage of the existing skeleton building algorithms is the loss of information about the width of the original figure sections, which is often necessary in image recognition and description tasks. The paper proposes an approach that follows the previously published method of skeletalization based on heuristic rules and consists in the sequential cutting off of figure segments with minimal chords in places where the border of the figure has a negative inflection when it is counterclockwise. Then segments are constructed connecting the midpoints of the chords of adjacent segments. The seg-ments are combined into chains that form a nonconvex figure skeleton. In this case, the lengths of the obtained chords carry information about a figure width in the corresponding sections. The experiments were related to two subject areas: processing scanned archival drawings of parts of a general engineering application to use previously gained experience in designing new products and reducing the overall design time and technological preparation of production, as well as the problem of recognizing a continuous handwritten text in the off-line mode.

◄ ← Preview | 1 | 2 | 3