| Summary: | Background. An analysis of existing methods for automatically searching for correspondence in stereo
images has shown that there are problematic issues in the field of automating the search for correspondence in stereo
pairs from the point of view of simultaneously minimizing computational costs and ensuring the accuracy of calculations.
They are associated, first of all, with the automatic determination of the position of the search area, limited on both sides and adaptive in shape and nature of distortion, under the conditions of ensuring the possibility of its restructuring
in real time with dynamic changes in the relative position in the space of video channels with known and unknown
internal parameters, and also when processing high-resolution images without their preliminary pixel-by-pixel processing.
Materials and methods. The problem to be solved by the proposed methods for automatically determining the position of
the search area for correspondence was to construct a search area for an object in a stereo image that is limited on both
sides and adaptive in shape by the nature of the distortion, provided that it is possible to quickly rearrange it in cases of
dynamic changes in the relative position in the space of digital images. video channels with known and unknown internal
parameters. As a result, two methods were developed for automatically determining the boundaries and position of the
adaptive search area, based on calculating the values of arrays containing the pixel coordinates of the object search area in
a stereo image. The adaptive property of the search area in this case consisted in the automatic change of its shape and size
in accordance with the geometric distortions of the images and the current relative position of the video channels. Results.
Increased performance compared to existing methods for determining areas of local search for correspondence is achieved
due to the absence of the need for preliminary pixel-by-pixel image processing, calculation of the fundamental matrix and
solving complex systems of equations. Conclusions. As a result, the requirements for computer computing resources have
been reduced and the real-time automatic search for matches in distorted images from high-resolution video channels dynamically
changing their spatial position and orientation has been achieved. In addition, the implementation of a technical
vision system is simplified due to the possibility of using video channels with unknown technical parameters.
|
|---|