Infrared Thermal Hand Vein Pattern Recognition

Ninth LACCEI Latin American and Caribbean Conference (LACCEI 2011), Engineering for a Smart Planet, Innovation, Information Technology and Computational Tools for Sustainable Development, August 3-5, 2011, Medellín, Colombia. Infrared Thermal Hand Vein Pattern Recognition Carlos Cabrera, Magno R. Guillen, Malek Adjouadi Florida International University, Center for Advance Technology and Education, Miami, Fl, USA Abstract An approach to biometric human authentication using thermal images of dorsa vein patterns is presented in this paper. In our work, an infrared (IR) camera is adopted as the input device to acquire the thermal images of the human dorsa. In this proposed approach, two thermal images are obtained; one in low resolution and the other in high resolution. Specific region of interest (ROI) on the thermal images are obtained with the support of an algorithm in MATLAB designed by the authors. Once the ROI is defined, feature points of the dorsal vein patterns (FPDVPs) are extracted based on the Fourier s law on heat conduction. Thus multi-resolution representations of images with FPDVPs are obtained through the implementation of multiple multiresolution filters (MRFs), which expose the dominant points by extracting miscellaneous features for each FPDVP. Keywords: Biometrics, Authentication, Thermal Image, Palm Veins. 1. Introduction Personal authentication has become both an imperative and high-demand practice for security access systems. Long-established conventional personal verification methods (PVM), such as passwords, personal identification numbers (PINs), magnetic strip cards, keys, and integrated circuit card (ICC) offer only restricted security and are untrustworthy. For example, cards or keys may be lost, stolen or just forgotten for an elapse period of time. Moreover, PINs may be easily obtained by any unauthorized person. Luckily, for the human race, our physical and physiological features acquire unique principals, such as universality, uniqueness, permanence, collectability, acceptability, and circumvention.[1] In order to prevent security issues intrinsic in long-established personal authentication methods, biometric authentication techniques have been intensively studied and developed to enhance the dependability of personal verification. All biometric verification techniques deal with various physical and physiological human features that include but not limited to fingerprints, dorsal geometry, handwritten signatures, retinal patterns, and facial images. In addition, the use of infrared (IR) images of biometric features, and the subcutaneous vascular network of the palm have furthermore been researched. Factors that may influence the user-friendliness and the attractiveness, as well as applicability and performance of biometric verification techniques are uniqueness, repeatability, maximum throughput, immunity from forgery, successful identification of dark-skinned subjects, false rejection rate (FRR) and false acceptance rate (FAR), ease of use (commonly referred to has user friendly). Even in today s world, there is not one biometric verification method or process that can fulfill all these requirements. In this paper, we will present a different approach for the authentication process augmented on the human vein patterns of the dorsa. In our current work, we are developing [2]an authentication system based on the use of infrared thermal vein patterns on the dorsal part of the hand. The physical figure of the subcutaneous vascular network of the back of the hand contains information that is able to authenticate the identity of any individual. The major purpose of this Medellín, Colombia WE1-1 August 3-5, 2011

research is to use thermal image processing methods to develop and implement real time algorithms capable of segmentation, thinning and feature extraction. Due to the human body vascular system, veins permeate the entire body. However, the vein patterns on dorsal part of the hand are particularly visible, easy to obtain and process. Anatomically veins are the blood carrying vessels inter weaved with the muscles and the bones. The prime function of the vascular (veins and arteries) system is to supply oxygen to the different body parts. The spatial arrangement of the vascular network in the human body is stable and unique and hence the patterns of veins are unique to every individual, even between the identical twins. 2 Related Work In recent times, several private and governmental agencies have approved out research into vein pattern biometric technology. Most of this research falls into the category of the vein patterns either in the face or hand. For facial vein patterns, the trait of the face is captured using a thermal camera that records the temperature of the vascular network, and hence is often referred to as facial [3]thermogram. From a face thermogram, there are abundant features that can be chosen for categorization. The most important ones are the thermal curves produced by the main arteries located in the human face. There are numerous thermal curves with diverse shapes that can be observed in a face. Personal authentication has been attempted by measuring the differentiation of many of these elemental. Compared to facial thermograms, acquiring the vascular human vein patterns of the hand is considered to be less covert and more user friendly. 3 Infrared Imaging of Hand Vein Patterns 3.1 MIR (describe acronym MIR mid infrared) Imaging Most all objects emit one form of radiation when subject to heat. The MIR imaging technology creates an image basically using the infrared radiation emitted by the human body. 3.1.1 Principle of Imaging: All objects radiate a constant range of frequencies. The total emissive power µ is described by the Stefan 8 2 4 Boltzmann law given in (1), where ξ is the emissivity of the object and α = 5.6x10 watt / m K is Stefan s constant and T is the temperature in Kelvin. The relationship between the wavelength λ max and black body temperature T is formulated by Wien s displacement law based on Planck s energy distribution law as given in (2) ω ξ σ λ 4 = x x T Eq. 1 3 = 2.9x10 / T Eq. 2 Characteristically, the human body emits infrared energy with λ in the range of 3 14 mm. These infrared waves radiate into the atmosphere and are attenuated according to the infrared transmittance spectrum of the atmosphere, and at the ranges of 3 5 and 8 14 mm, the radiant emittance of infrared spectrum possesses the highest transmittance rate. Therefore by using a thermal camera with detector sensitivity in the range of either 3 5 or 8 14 mm, an image showing the heat distribution of the human body can be generated. Medical researchers have found that human superficial veins have higher temperature than the surrounding tissues. Therefore via thermal imaging, the images containing the heat distribution of body parts can clearly display the structure of the desired vein patterns, as can be seen in Fig. 1. MIR imaging forms images using the infrared radiation emitted from the objects. No external lighting is required. Therefore MIR imaging does not suffer from illumination problems like many other imaging techniques. Medellín, Colombia WE1-2 August 3-5, 2011

Fig. 1 MIR images taken from the same hand under room temperature. 4 Image Acquisitions To capture the thermal images we used mid-wavelength infrared (MWIR) high performance camera, consisting of a Stirling-cooled Indium Antimonide Focal Plane Array (FPA). The FPA is a 320 x 256 array of detectors that are sensitive in the 1.0 to 5.4 µm range. ThermaCAM Researcher software is being used in support of the imaging algorithm that is currently being developed. This software is used to view, measure, and analyze sequences of IR images and thermal events. The measurements are made using the following tools: isotherm, spotmeter for areas and lines. The entire system will be integrated using a MATLAB graphical end-user interface (GEUI) to provide a user-friendly environment. In MATLAB, the following design was created using SIMULINK. The purpose of this network is to process low resolution images and filter out noise that could have been acquired during input. This process is called: Noise Reduction. This design calls for the use of 2D filter. 4.1 Image Processing Fig. 2 MathLab\Simulink design of a 2D filter. Once the thermal image has been acquired and filtered, a process of image segmentation for vascular extraction begins. This is an iterative process. In our research, it has been found that the optimal number of iterations should be no less than 100. Contrary to other published research, it has been observed that even utilizing low resolution images acceptable results are obtained. Moreover, using high resolution also gives us higher noise, given that all frequencies are amplified. Fig. 3 depicts a low thermal resolution while Fig. 4 depicts a high thermal resolution image. Both images, from a dorsal vascular point of view, give us the same information, although Fig. 4 has high level of noise, which have been circled. Noise can be observed in the form of dorsal hair, thermal boundaries and even the thermal shock effect around the skin. This is produced from the thermal difference in room temperature at 75º F (23.8º C) and human temperature at 98.6º F (36º C).[4, 5] In order to filter the noise in these thermal images, a 2D filter circuit had to be designed and implemented. We must keep in mind that filtering is a process of selecting Medellín, Colombia WE1-3 August 3-5, 2011

frequency components from a signal. By inspection, it can be seen from that the MIR imaging technique can capture the major vein patterns in the dorsal of the hand. As a replacement for of our discussion in regards for 1D signal that characterize changes in amplitude in time, at this point we are working with 2D signals which characterize intensity variations in space. We achieve these images really has thermal images. The images we process here have the following characteristics: first they are digital, and therefore have a finite area given by width and height and are measured in pixels, and are fixed in amount. Given that the thermal signals are discrete, we must obtain the analog component of the 1 dimensional DFT 1. This component is given by the following FFT transforms: Eq. 3 Accordingly, an M x N thermal image has an M x N matrix set of complex Fourier coefficients. To implement this transform, we would like an analog of the FFT 2 [6]. This will allow us to compute the FFT coefficients using less computational power. In reality, we have achieved further. We know that the 2 Dimensional Discrete Fourier Transform is divisible into two, one dimensional Discrete Fourier Transform which can be executed with a simple FFT algorithm governed by the following equation: Eq. 4 Eq. 5 However, MIR 3 thermal imaging of vascular patterns presents the shortcoming that marks and imperfection on the skin surface are also visible in the thermal image. This issue will damage the network of the vein patterns and result in problems when it comes to further image processing and pattern recognition stages. Fig. 3 Fig 4 Once we have captured and filtered out dorsal trait, the area of segmentations must be defined. Unfortunately in this process this is done manually. However, efforts in our research are in the way to make this autonomous. The 1 DFT: Discrete Fourier Transform 2 FFT: Fast Fourier Transform 3 MIR: Mid Range Infra-Red Medellín, Colombia WE1-4 August 3-5, 2011

entire dorsal of the hand has been selected as shown in Fig 5. After 100 iterations, it has been clearly outlined the contour of our trait (ROI) as shown in Fig 6. Fig. 7 depicts the code in MATLAB for the iterations. m = imresize(m,1); % for fast computation = 1:NumofFiles Fig. 5 Fig. 6 Fig. 7 p = floor(i /1000); q = floor(i / 100); r = floor(i/10); % I = imread(strcat(pathname,num2str(p),num2str(q),num2str(r),num2str(i),'.jpg')); I = load(strcat(pathname, Filename)); % I = double(i); figure; [outcellarray] = structtocellarraywithheaders (I); I = outcellarray(2,1); I = I[6]; imshow(i,[]); %m = false(size(i,1),size(i,2)); %-- create initial mask m = roipoly; %m(123:152,121:151) = 1; m = uint8(m); %m(37:213,89:227) = true; I = imresize(i,1); %-- make image smaller for i Since our veins grow in the same proportion as the human been grows, then the pattern associated to them never changes. This pattern is built-in at conception[7]. Only the outline of the vein pattern can be used as the sole feature to distinguish each individual. A good illustration of the pattern s shape is via extracting its skeleton. Fig. 8 shows the skeleton of the vein pattern after applying the thinning algorithm proposed by the authors. It can be observed that after implementing the vein smoothing process, the skeletons of the vascular pattern are effectively extracted and the shape of the vein pattern is well observed. However, some noise does appear in our final result. Fig. 8 It is noticeable the contrast between the first thermal images and the final result obtained, as shown in Figs. 9 and 10: Fig. 9 Fig. 10 Medellín, Colombia WE1-5 August 3-5, 2011

5. Conclusion The central problem in biometric authentication is to find an efficient and effective approach which can characterize biometric features and measure the extent of similarity or distinction among two or more persons. In this paper, a novel approach based on the thermal vein patterns of dorsa was presented. Authors are also conscientious that although much has been progressed more research is needed. Since the demand for information, network bandwidth, computer processing power and personal security increase exponentially every year, faster and real time algorithms for authentication are needed. 6. Acknowledgment The author would like to acknowledge the support of Dr. Malek Adjouadi, Director of the Center for Advance Technology and Education, (CATE) at Florida International University, Miami, Florida, U.S.A., and of Dr. Magno Guillen, Senior Neuro Data Imaging Analyst at Miami Children Hospital, Miami, Florida, U.S.A., for their unconditional support and endorsement provided to carry and continue to carry this research. 7. Future Work The breakthrough we have achieved is just the beginning; the authors are currently working in developing the algorithms to simulate the new system stated. We will extract the statistical results to determine the system response in the worst case scenarios. 8. References 1. M. K. Shahin, A.M.B., M. E. Rasmy, A MULTIMODAL HAND VEIN, HAND GEOMETRY, AND FINGERPRINT PROTOTYPE DESIGN FOR HIGH SECURITY BIOMETRICS. CIBEC'08, 2008. 1. 2. Jianjun Zhao, H.T., Weixing Xu and Xin Li, A New Approach to Hand Vein Image Enhancement. Second International Conference on Intelligent Computation Technology and Automation, 2009. 1: p. 3. 3. Gnee, N.S., A Study of Hand Vein, Neck Vein and Arm Vein Extraction for Authentication. ICICS 2009. 1. 4. Lovell, A.K.M.H.V.K.M.B.C., Biometric Authentication based on Infrared Thermal Hand Vein Patterns. Digital Image Computing: Techniques and Applications, 2009. 1(1): p. 8. 5. Chih-Lung Lin and Kuo-Chin Fan, M., Biometric Verification Using Thermal Images of Palm-Dorsa Vein Patterns. IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, 2004. 14(2): p. 15. 6. Wongsawat, Y.P.a.Y., Palmprint Image Enhancement Using Phase Congruency. IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY International Conference on Robotics and Biomimetics, 2008. 1: p. 4. 7. David Zhang, G.L., Wei Li,, Palmprint Recognition Using 3-D Information. IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS, 2009. 39(5): p. 15. Authorization and Disclaimer Authors authorize LACCEI to publish the paper in the conference proceedings. Neither LACCEI nor the editors are responsible either for the content or for the implications of what is expressed in the paper. Medellín, Colombia WE1-6 August 3-5, 2011