DEVELOPMENT OF A METHOD FOR SYNTHESIZING AN APERTURE DIAPHRAGM IN HYPERSPECTRAL REMOTE SENSING SYSTEMS FOR EARTH

The aim of the work is to research and develop methods for synthesizing aperture in hyperspectral systems for remote sensing of the Earth to reduce weight and size characteristics by applying methods of program-algorithmic processing of the input signal and implementing the synthesized aperture. A method of neural networks for deconvolution on the construction of a radial basis network is developed. A method has been developed to increase the synthesis of apertures in hyperspectral systems for remote sensing of the Earth. A method for increasing the spatial resolution of images obtained by optical systems for remote sensing of the Earth is described. A method for radiometric calibration of output data has been developed, which allows universalizing the analysis of spectral characteristics. In the process, to achieve the goals were used: methods of spectral optics, mathematical analysis and statistics, methods of processing images and signals. The project results contribute to the reduction of overall weight and cost characteristics and the possibility of synthesizing the aperture at the exit of the polychromator, which will avoid the use of expensive camera lenses in hyperspectral systems of remote sensing of the Earth. The developed methods for synthesizing aperture in hyperspectral systems of remote sensing of the Earth differ from the existing ones in that the receiving device for the video signal does not contain structural changes, and they contain the function of the algorithmic apparatus, which includes the analysis of the functions of the scattering point, the deconvolution of the recorded signal is performed by the method of neural networks after the stage learning.

Optimization of synchronous reluctance motor based on radial basis network

This paper presents surrogate-model based optimization for synchronous reluctance motor (SynRm) with transversally laminated rotor. A radial basis function (RBF) model with 12 input variables and three outputs is first trained. A dataset is obtained using finite element method to estimate parameters of RBF model. By building RBF model, the RBF network can predicts the outputs of the SynRm with good accuracy Using non-dominated sorting genetic algorithm (NSGA II), pareto front is obtained. The SynRm is designed to maximize the maximum developed torque and power factor of the motor with constrained torque ripple.