Thermal Biometric Features for Drunk Person Identification Using Multi-Frame Imagery

In this work, multi-frame thermal imagery of the face a person is employed for drunk identification. Regions with almost constant temperature on sober and persons are thoroughly examined their capability to discriminate intoxication. Novel image processing approaches as well feature extraction techniques developed support identification procedure. These constitute novel ideas in theory analysis algorithm development. Nonlinear anisotropic diffusion light smoothing images before extraction. Feature vector based morphological operations performed isothermal regions face. The classifier chosen verify discrimination capabilities procedure Support Vector Machine (SVM). Obviously, change shape size alcohol consumption. Consequently, intoxication can be carried out only signatures person, while signature corresponding not needed. A sample 41 participants who drank controlled consumption was creating database, which contains 4100 images. proposed method achieves success rate over 86% constitutes fast non-invasive test that replace existing breathalyzer check up.