METHOD OF RECOGNIZING OBJECTS ON AN IMAGE Russian patent published in 2019 - IPC G06T7/00 

FIELD: computer equipment.

SUBSTANCE: present invention relates to methods of processing visual information and can be used in vision systems when solving problems of searching, tracking, guidance, diagnosis, control and recognition of objects on an image. Method comprises dividing an image into zones and selecting structural elements, wherein the image is divided into zones by filtering objects on the binary contour image based on the eight-coupling principle. Subsequent selection of structural elements is carried out by determining significant points for each selected object – end points, intersection points and bends of lines, after that, information on parameters of primary links of each pair of selected significant points is stored. A selection of rectangular objects and simple closed figures on the image is performed by analyzing the stored information on parameters of primary links of each pair of selected significant points, method includes storing information obtained from analysis on secondary links – relationships between each pair of significant points of selected objects.

EFFECT: high accuracy and reliability of recognizing objects on an image in conditions of different illumination and changing weather conditions of photography.

1 cl, 4 dwg