Pattern recognition of thermovision images for biometrics

The main goal of this work is the automatization of recognition process and proving, whether the thermal imaging can be exploited for realization of the biometric system. The data base was created of the thermal images of human faces. The system was designed for realizing 3 various functions: verification – testing whether the person to be verified is in data base, authorization – testing whether the pictures of the person is placed in samples data base, and identification – system performs the comparison of the actual image with these in data base (all images of the person are searched. The optical recognition methods 4f correlator and joint transform correlator were applied and compared with optoelectronics and computing methods, statistical Eigenfaces and Fisherfaces analysis, and Multilayer Perceptron - neural network. As the result we have stated that the thermal image is a biometric feature, people can be efficiently recognized based on thermal images, and the system efficacy depends on the proper selection of recognition method.