| CPC G06V 10/74 (2022.01) [G06V 20/10 (2022.01); G06V 20/35 (2022.01)] | 13 Claims |
|
1. An operation method of a server for identifying disaster affected areas by using disaster images with machine learnings, the operation method comprising:
acquiring at least one first disaster image;
deriving a first area from the at least one first disaster image, and acquiring first area related information through labeling based on the derived first area;
training a first learning model using the at least one first disaster image and the first area related information;
deriving a plurality of second areas based on the derived first area in the at least one first disaster image; and
acquiring second area related information for each of the plurality of derived second areas based on labeling,
wherein
the first learning model is trained further using the second area related information for each of the plurality of derived second areas,
the second area related information for each of the plurality of derived second areas includes feature information for each of the plurality of second areas, and
the feature information is information which is set considering a relationship between each of the plurality of second areas and the first area;
acquiring a second disaster image from an external device;
deriving a first area and a plurality of second areas from the second disaster image, and acquiring a plurality of disaster related information through labeling based on the derived first area and the derived second areas;
assigning a weight to each of the plurality of acquired disaster related information; and
inputting the second disaster image and the plurality of disaster related information to the trained first learning model; and outputting first area identification information and disaster damage type information based on the first learning model,
wherein
the second disaster image is inputted to a second learning model,
the second learning model derives the first area and the second areas of the second disaster image, and provides the plurality of disaster related information as output information through the labeling based on the derived first area and the derived second areas,
the first area is an area which a disaster affects, and the second areas are areas which are located within a predetermined distance from the first area,
the plurality of disaster related information generated through labeling in relation to the first and second disaster images include affected area information, damage type information, neighborhood feature information, damage propagation information and weather information at the time of the disaster, and
the labeling of the first and second disaster images is performed in relation to the first area, and the plurality of disaster related information is generated through the labeling based on the second areas.
|