METHOD AND APPARATUS FOR IDENTIFYING PORTIONS OF FRAGMENTED MATERIAL WITHIN AN IMAGE Russian patent published in 2019 - IPC G06T1/40 G06K9/46 G06K9/62 G06T7/13 G06N3/04 G06N3/08 

FIELD: image processing means.

SUBSTANCE: invention relates to image processing. Technical result is achieved by the fact that the method includes receiving pixel data corresponding to an input plurality of pixels representing a fragmented material image, processing pixel data using a convolutional neural network having a plurality of layers and obtaining an output representation of the pixel classification, indicating whether the pixels are in the input pixel set in one of the following positions: on edge of fragmented material part, from inside from edge and in gaps between parts of fragmented material. Convolutional neural network comprises at least one convolution layer configured to generate a convolution of an input plurality of pixels, and is preliminarily trained using a plurality of training images containing previously identified portions of fragmented material. Method further includes processing an output representation of the pixel classification in order to match the identified edges with parts of the fragmented material.

EFFECT: identification of parts of fragmented material within an image.

28 cl, 19 dwg