Nonparametric assessment of two-dimensional density of nonlinear location signal distribution for object recognition despite background noise

Abstract

Problem statement Searching for objects in shelter environments is a challenge associated with the difficulty of obtaining and interpreting search results. Currently, within the statistical theory of detection and recognition, no models and methods have been developed that allow automatically identifying the detected object. This is due to the difficulty of describing signal and noise models in the form of known statistical distributions. The variety of signals leads to the inability to describe them by any universal law of distribution. Therefore, traditional detection and recognition criteria do not provide high reliability of object search and identification. The purpose of the study. To develop and test the method for nonparametric estimation of nonlinear location signals, assess the potential possibilities of applying the developed method for automatic (automated) signal detection (recognition) systems. Results. The model for recognizing nonlinear location signals based on nonparametric approximation of two-dimensional probability distribution densities is presented. The potential possibility of improving the efficiency of nonlinear locators is shown. Practical significance. For nonlinear location systems a toolkit for estimating the probability of the detected object belonging to a given class of objects is proposed.