Publication of the journal

The section is currently being updated

The existing recognition methods for vision systems are effective only for certain fairly simple objects, which are observed in sufficiently deterministic conditions (certain illumination, background and position of the object relative to the camera). During functioning of robotic complexes on a previously unprepared territory, the listed conditions, as a rule, are not fulfilled. Thus, the prerequisites for the implementation in perspective the robotic complexes of the algorithms for recognizing the three-dimensional objects of various classes in the complex nondeterministic conditions are created. The issues of recognition of three-dimensional objects with an unknown viewing angle from the images of their two-dimensional projections have been considered. The multihypothetical recognition is based on the generalization of a two-alternative decision rule. The questions of using the ROC-analysis for evaluating the effectiveness of the recognition system have been investigated. The formula dependencies to determine the magnitudes of possible errors have been presented. The obtained results can be used in the development of methods and ways for identifying objects in conditions of an inadequately defined environment for vision systems of robotic complexes for various purposes operating outside the factory premises.

124498, Moscow, Zelenograd, Bld. 1, Shokin Square, MIET, editorial office of the Journal "Proceedings of Universities. Electronics", room 7231

+7 (499) 734-62-05
magazine@miee.ru