Article 2114

Title of the article



Terekhin Andrey Viktorovich, Postgraduate student, Murom Institute (branch) of Vladimir State University
(23 Orlovskaya street, Murom, Vladimir region, Russia),

Index UDK



Background. Intensive development of modern technical provisions in the field ofproduction control, improvement of the computing power of modern computers, high quality requirements, intense competition, enterprise’s needs in improving the quality and production rates make the problem industrial automation a topical one. A variety of computer vision systems (CVS) does not exclude the fact that all of them are designed for specific tasks. All of them use different attributes and implement various detection algorithms. The task of identifying three-dimensional objects is still a fairly new one, and tools and algorithms of planar geometry are still used for its solutions. In this regard, the development of new three-dimensional object recognition algorithms is still topical. The article describes a new approach to the recognition of three-dimensional objects in two images by calculating the estimates using the diagonal form features and octree models.
Materials and methods. For the research the author designed a special program for a three-dimensional object recognition computer. To test objects, the generated orthogonal and oblique projections of the objects were sent to the progrma input. There were 1500 tests – each of 60 projections was generated 25 times (10 objects with 6 projections). The estimate calculation algorithm was used as a recognition algorithm. Before calculating the diagonal form features, images were preprocessed (filtering, binarization, edge detection).
Results. The research includes determination of the influence of a number of features used for recognition on probability of correct recognition, and the influence of using oblique projection on probability of correct recognition in general. 100% average probability of correct recognition of three-dimensional objects in two images using the entire set of features was experimentally proved.
Conclusions. The proposed approach could be used in computer vision systems for assembly conveyors where it is necessary to recognize three-dimensional objects that are randomly arranged even in situations when some of them might have similar projection.

Key words

recognition, three-dimensional object features of shape, octree, CVS.

Download PDF

1. Bubennikov A. V. Nachertatel'naya geometriya: ucheb. dlya vuzov [Descriptive geometry: textbook for universities]. Moscow: Vyssh. shk., 1985, 288 p.
2. Sadykov S. S., Savicheva S. V. Priborostroenie [Instrument engineering]. 2012, no. 2, pp. 19–24.
3. Terekhin A. V. Nauka i sovremennost' – 2013: sb. materialov XX Mezhdunar. nauch.-prakt. konf. (Novosibirsk, 20 fevralya 2013 g.) [Science and modern times – 2013: proceedings of XX International scientific and practical conference (Novosibirsk, 20 February 2013)]. Ed. S. S. Chesnokov. Novosibirsk, 2013, 300 p.
4. Zhuravlev Yu. I., Nikiforov V. V. Kibernetika [Cybernetics]. 1971, no. 3, pp. 1–11.
5. Jackins C. L., Tanimoto S. L. CGIP. 1980, vol. 14, pp. 249–270.
6. Noborio H., Fukuda S., Arimoto S. IEEE Trans. PAMI. 1988, vol. 10, no. 6, pp. 769–782.
7. Chien C. H., Aggarwal J. K. CGIP. 1986, vol. 36, pp. 100–113


Дата создания: 28.08.2014 09:39
Дата обновления: 28.08.2014 09:58