DEVICE AND METHOD FOR RECOGNIZING CONTOURS OF AGRICULTURAL FIELDS USING DEEP LEARNING BASED ON REMOTE SENSING DATA Russian patent published in 2022 - IPC G06T7/12 G06V10/762 

FIELD: computer technology.

SUBSTANCE: technical solution relates to the field of computer technology. The expected result is achieved due to the stages at which: a multispectral image is obtained; a segmentation mask for the arable land class, a segmentation mask for the agricultural field boundary class and a mask of distances from the center of the agricultural field to the nearest border of this field are determined for the image; a topological mask is formed that determines the presence of arable land on a multispectral image; based on the values of the mask of distances from the center of the agricultural field to the nearest border of this field, the starting points are determined; the connected clusters of the starting points are marked with separate identifiers; the points of the multispectral image are distributed over the clusters defined above from the connected clusters of the starting points defined above using the values of the mask of distances from the center of the agricultural field to the nearest border of the field; clusters of points connected separately by a single identifier are obtained; the contours of agricultural fields as the coordinates of the reference points of polygons forming the geometry of individual agricultural fields are defined.

EFFECT: increase in the accuracy of determining the contour of the agricultural field.

8 cl, 11 dwg