EVALUATION OF THE DISTANCE TO THERMAL OBJECTS BY PASSIVE MEANS OF THE MID-WAVELENGTH INFRARED BAND: A MULTIDIMENSIONAL REGRESSION APPROACH
A. V. Kvasnov
St. Petersburg Federal Research Center, Russian Academy of Sciences, St. Petersburg, Russia
Keywords: infrared band, thermal camera, multilayer perceptron, regression analysis, radiation transfer
Abstract
The article considers an approach to range determine to point targets based on the images from a thermal camera in the mid-infrared band. The paper shows the analysis of the equation of radiation transfer and demonstrates the statistical stability of the target thermal distribution depending on the range and weather conditions. A multidimensional linear regression model is proposed based on the identified object features in order to predict the distance to the object using a trained artificial neural network. The article simulates the obtained approach according to known experimental data, which showed the advantage of the estimated distance (less than ΔD/D < 0,04)) in comparison with the direct method of the range determining (ΔD/D > 0,82). The ways forward of research are given and the need to improve the feature space is valid for the increasing of the accuracy.
|
