Analysis of the effectiveness of object recognition algorithms using a thermal imaging camera

The work presents a comparison of the object recognition methods based on three algorithms (Eigenface, Fisherfaces, LBPH - Local Binary Patterns Histograms), which were extended to include thermal imaging. The aim of the work is to select the most effective analytical method in changing environmental conditions and implement it in the form of software for a technological project called OctoView. The system placed on a drone, combined with efficient processing units and Jetson NANO microcomputers in cooperation with a multi-vision system, will provide comprehensive analysis and image assessment of “threats” located in the observed area, recording field events, tactical risk analysis and reports, monitoring areas border areas, status and position and type of own resources in border (or other) sections. This project is dedicated to military, border, customs, fire brigade and police applications. The purpose of implementing the software is to increase security in border areas through continuous image observation and analysis [1]–[3], as well as monitoring threats and their dynamics.