METHOD OF SELECTING A FEATURE VECTOR FOR RECOGNIZING IMAGES OF OBJECTS Russian patent published in 2019 - IPC G06K9/46 

FIELD: image processing means.

SUBSTANCE: invention relates to image processing. A method of extracting a feature vector for recognizing images of objects involves converting input signals into electrical video signals, sampling said signals, splitting the image into discrete elements. Scanning of the image is performed with an electronic window; the selection of the image vector of the images of the objects is performed from images. Selection of the object images feature vector for each fragment of the decrypted image limited by the electronic window is performed by forming a vector of discrepancies between the ordered values of the sample of intensity of video signals of the analyzed fragment of the decrypted image and ordered by magnitude values of video signal strengths of well-formed reference sample of the same size for image of recognized objects, discrepancy vector norm is calculated and its value is compared to the value of the threshold to be set.

EFFECT: high versatility of the method of selecting a feature vector, low resource intensity, high efficiency of object recognition, high probability of recognizing objects with fuzzy boundaries, reduced sensitivity to changing the scale of images and angular orientation of objects on the image, as well as improved functional capabilities of the method, which include recognition of objects based on data on parts thereof.

1 cl, 3 dwg