APPLICATION OF R-CNN IN AUTOMATIC IN-DOOR POSITIONING THROUGH ANNS ANALYSIS OF GRAPHICAL DATA

Abstract

The most relevant approaches used in the framework of the construction of optical monitoring and automatic control systems based on video registration data arrays by applying neural network analysis algorithms have been determined. The advantages of using convolutional neural network architectures for solving this class of problems, which perform analysis based on the definition of areas of interest, as well as the method of support vectors and regression models, are indicated. The purpose of the machine analysis relevant algorithms optimization concept development was to reduce the load on the computing resource in the process of selection and classification of the visual object through the application of the procedure of selective search of regions of interest for each image matrix and subsequent determination of the feature vector for each region of interest. The developed mathematical apparatus can be effectively used in solving a wide class of geolocation problems and provides an opportunity to evaluate the optimization of the machine analysis system in accordance with the target accuracy indicators, reducing the processing time of the incoming request and reducing the load on the computing resource of the hardware complex.