Intelligent MRTD Testing For Thermal Imaging System Using ANN

One of the most important parameters of a thermal (or infrared) imaging system is the Minimum Resolvable Temperature Difference (MRTD). It brings in the performance of each subsystem and correlates well with system's effectiveness under operational conditions. So it gives a more comprehensive characterization of infrared imaging system performance. The MRTD testing is a psychophysical task. Many subjective and objective models have been proposed to predict it. The intelligent MRTD testing has been realized because of the development of neural network. The network can partly simulate the brains composition and ability. So the network trained by subjective measurement results can have the human's judgement. Using ANN, computer can extract features recognized by ANN from the grabbed thermal images, and then give results conforming to the human judgement. This method can reduce the uncertainties between observers and long time dependent factors by standardization. Especially when measuring imagers of the same type, the test are time and cost saving, and can reduce fluctuation of the results.This paper will introduce the intelligent MRTD testing models, realized method and demonstrate the feasibility. The feature extraction algorithm, model of neural network and analysis of results are described briefly. The intelligent MRTD testing is based on the subjective training data, so the factors due to the subjective nature of the observers also influenced the accuracy of intelligent testing. Considering the errors of the arithmetic and model, the main factors are discussed in detail in order to decrease the influence on the accuracy of intelligent testing. At the same time, the signification of intelligent MRTD testing and the problems need further research are also mentioned.