In this work the statistical pattern recognition methods were applied for evaluation of transillumination images of interphalangeal joints of patients suffering from rheumatoid arthritis. Special portable apparatus was constructed for performing the transillumination examination. It consisted of He-Ne laser with optics for collimated illumination, special holder for placing the finger (perpendicular to optical axis, dorsal site towards camera), and CCD camera with memory stick. 20 ill patients and 20 healthy volunteers were examined. The captured images with 1152x864 resolution were converted into the gray level pictures and analyzed by means of statistical pattern recognition method. Principal Component Analysis (PCA) and cluster analysis by use of 1-NN method (1 Nearest Neighbour) were applied for classification. The recognition system was able to differentiate correctly between healthy and ill subjects with 72.35% accuracy in case the data base composed of 40 persons.
Human body posses many unique, singular features that are impossible to copy or forge. Nowadays, to establish and to ensure the public security requires specially designed devices and systems. Biometrics is a field of science and technology, exploiting human body characteristics for people recognition. It identifies the most characteristic and unique ones in order to design and construct systems capable to recognize people. In this paper some overview is given, presenting the achievements in biometrics. The verification and identification process is explained, along with the way of evaluation of biometric recognition systems. The most frequently human biometrics used in practice are shortly presented, including fingerprints, facial imaging (including thermal characteristic), hand geometry and iris patterns.
In this paper, the application of the optical correlators for face thermograms recognition is described. The thermograms were colleted from 27 individuals. For each person 10 pictures in different conditions were recorded and the data base composed of 270 images was prepared. Two biometric systems based on joint transform correlator and 4f correlator were built. Each system was designed for realizing two various tasks: verification and identification. The recognition systems were tested and evaluated according to the Face Recognition Vendor Tests (FRVT).
The distribution of light emitted from the sol-gel coated fiberoptic applicator for interstitial laser therapy was examined. The silica sol-gel coatings were prepared from silicate precursor TEOS (tetraethylorthosilicate) mixed with ethyl alcohol in acid catalyzed hydrolysis. The applicator was produced with ratio R = 20, whereas R denotes the number of solvent moles (ethanol) to the number of TEOS moles. Then, the light pattern was captured by means of CCD camera and statistical pattern recognition method was applied to study the repeatability of light distribution. Two methods were applied: Fisher Linear Discriminant Analysis (FLDA) and Principal Component Analysis (PCA). The top match characteristic TM as calculated by statistical pattern recognition methods was taken as a factor corresponding to the performance repeatability. It was demonstrated that is possible to produce the sol-gel applicators with more than 92% repeatability.
Biometrics is recently a key interest of many institutions, especially those interested in public security. Biometrics relies on recognition, identification and verification of physiological or/and behavioral features of human body. These include e.g. fingerprints, retina or iris images, examination of DNA or way of handwriting and many others. Face recognition is one of these modalities. The successful security systems should offer high efficiency, meaning high rate of properly classified and verified images. They should be also free from external disturbances like unstable environment conditions, illumination etc. In this paper we demonstrated that biometric verification based on the analysis of thermal images could fulfill these requirements and thermal image of the human face can be regarded as a biometric feature, as well. Analyzed database incorporates thermal images of 26 adults, registered by high quality infrared camera AGEMA 900 LW in various conditions under the influence of externally induced changes to the thermal characteristics. In order to perform the image processing the compression step was applied, so the acquired 270 x 135 pixels image was reduced to the size of 85 x 85 pixels. The reduced images were further analyzed. Our system worked in three regimes, based on Eigenfaces, Fisherfaces and joint transform correlator. Receiver Operator Characteristic was used for evaluation of each of the mentioned above modalities. Equal Error Rate was applied for evaluation of the tested systems. It was shown that the highest rate of system efficiency was achieved by implementation of Fisherfaces 96,54% and joint transform correlator 88,40%. Thus, it was also demonstrated that temperature distribution can be regarded as biometrics feature and can be used for person verification.