Aerospace Instrument-Making Annotation << Back NEURAL NETWORK INTELLIGENT ALGORITHMS FOR SEGMENTATION OF HIDDEN SUBSURFACE OBJECTS OF REMOTE MONITORING BASED ON DATA OF DIFFERENT-TIME IMAGES IN THE VISIBLE AND INFRARED WAVELENGTH RANGE Yu.Yu. Gromov, I.N. Ishchuk, A.M. Filimonov, A.A. Zenkin The article describes and analyzes the operation of a neural network based on a three-layer Rosenblatt perceptron and a convolutional neural network U-net. Due to the need to increase the efficiency of segmentation and recognition of remote monitoring objects, including hidden subsurface objects, using multispectral optoelectronic systems of complexes with unmanned aerial vehicles and insufficient implementation of algorithms for processing multi-time intelligence data obtained in daylight and dark in automatic mode, with obtaining numerical estimates of the thermophysical parameters of hidden subsurface objects, ensuring their classification by types of structural materials, the application of a neural network intelligent algorithm in the problem of processing a cuboid of different-time infrared images and images in the visible wavelength range obtained using a multispectral optoelectronic system of an unmanned aerial vehicle is considered, which makes it possible to identify areas with similar dynamics of changes in thermal contrasts due to the corresponding thermal inertia of materials of hidden subsurface objects and backgrounds. The results of segmentation based on the proposed neural network algorithm are presented. As a result of processing such data, graphical visualization of spatial data related to the class of materials of objects and backgrounds becomes possible. Keywords: neural network, algorithm, optoelectronic system, unmanned aerial vehicle, object, segmentation. DOI: 10.25791/aviakosmos.3.2023.1329 Pp. 36-47.