METHOD FOR REMOTE ESTIMATION OF SPATIAL DISTRIBUTION OF THERMOPHYSICAL PARAMETERS OF OBJECTS AND BACKGROUNDS Russian patent published in 2021 - IPC G01J1/00 G01J5/00 G01V8/02 G01N25/18 G01N25/20 

FIELD: thermal control.

SUBSTANCE: invention relates to techniques for active non-destructive thermal control and can be used in equipment for remote sensing of the earth. According to the claimed method, the survey area is surveyed in the daytime in the visible and infrared ranges and in the dark in the infrared range. For images of the visible range, a databank of typical objects and backgrounds for remote monitoring is created, containing a table of values ​​of emissivity coefficients for each object of the classifier, measured in advance and recorded in the classifier table. For images of the infrared range, a databank of reference structural and functional materials of objects, anthropogenic and natural landscapes is created, containing a table of values ​​of thermal conductivity, thermal diffusivity, specific heat, density and thermal inertia for each object of the classifier, previously measured and recorded in the table. According to the images of the visible range, obtained during daylight hours, objects and backgrounds are classified based on algorithms of convolutional neural networks, their class is determined, and the values ​​of the emissivity at each point of the image are assigned from the corresponding data bank. The images of the infrared range, obtained in daylight and dark hours of the day, are converted into the spatial distribution of thermodynamic temperatures, which are classified on the basis of convolutional neural algorithms according to the thermophysical properties of structural and functional materials of objects, anthropogenic and natural landscapes at each point of the image with the assignment of tabular values ​​of thermal conductivity, thermal diffusivity , specific heat, density and thermal inertia from the corresponding data bank. An increase in the reliability of the classification of table materials is ensured by the possibility of adding thermodynamic temperatures recalculated from infrared images, as well as simulated infrared signatures of objects, to the training set, based on methods of deep learning of neural networks with a teacher.

EFFECT: increasing the reliability of determining the thermophysical parameters of remote monitoring objects, as well as simplifying the method for remotely determining the spatial distribution of thermophysical parameters of the earth's surface by eliminating the blocks for registering and issuing meteorological conditions and registering the amount of total solar radiation.

1 cl, 3 dwg