Gradient Approach to the Task of Contour Extraction of Big Anthropogenic Objects in the Images of Multi-Object Ground Scenes
| Autor: | S. V. Tikhonova, A. V. Rankova, D. A. Fortinsky, V. V. Insarov |
|---|---|
| Rok vydání: | 2015 |
| Předmět: |
Brightness
Relation (database) business.industry Computer science ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Image processing Bearing (navigation) Edge detection Computer Science Applications Image (mathematics) Human-Computer Interaction Reduction (complexity) Artificial Intelligence Control and Systems Engineering Digital image processing Computer vision Artificial intelligence Electrical and Electronic Engineering business Software |
| Zdroj: | MEHATRONIKA, AVTOMATIZACIA, UPRAVLENIE. 16:415-421 |
| ISSN: | 1684-6427 |
| DOI: | 10.17587/mau.16.415-421 |
| Popis: | The tasks of navigation and guidance of the unmanned maneuverable aircraft often involve a problem of detection and recognition of the land scene objects based on comparison of the priori and posteriori information (reference and observed images, respectively). Such a comparison allows us to localize the sought objects on the observed image and to define values of the current coordinates of these objects. The most stable features of the current image bearing information on the geometry of objects of an observed scene are the edges of the images of the objects of a scene, their placement and their direction. This information could be obtained from the brightness gradients. In this article the authors consider possible modification of the edges detection algorithm in relation to the solution of the problem of detection of contours of large technogenic objects in the image of a multi-object scene. This modification makes it possible to take into consideration the direction of the detected edges. The purpose of the proposed modification is reduction of the contours of the textures and small objects. Such edges often change their directions, respectively, they are broken into a number of small edges of different directions. Later such edges can be excluded from the consideration. At the same time the contours (edges) of large technogenic objects seldom change their directions. These contours bear the most useful and exact information about separate objects and a scene as a whole. The proposed edge detection algorithm was tested on the image of a real multi-object scene. Advantage of the proposed algorithm in comparison with the other known algorithms is demonstrated in application to the considered type of scenes. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|