Near- and Far- Infrared Imaging for Vein Pattern Biometrics

Transcription

1 Near- and Far- Infrared Imaging for Vein Pattern Biometrics Wang Lingyu Nanyang Technological University School of Computer Engineering N4-#2A-32 Nanyang Avenue, Singapore Graham Leedham UNSW Asia 1 Kay Siang Road, Singapore G.Leedham@unswasia.edu.sg Abstract This paper investigates two infrared imaging technologies, far-infrared thermography and near-infrared imaging, to acquire hand vein pattern images for biometric purposes. The imaging principles for both technologies are studied in depth. Experiments involving data acquisition from various parts of hand, including the back of the hand, palm, and wrist are described using a population of 150 participants using both near and far infrared imaging techniques. Comparison and analysis of the data collected show that far-infrared thermography has difficulties in capturing vein images in the palm, and wrist. However, while it is more suitable for capturing the large veins in the back of the hand, it is sensitive to ambient conditions and human body condition and does not provide a stable image quality. On the other hand, near-infrared imaging produces good quality images when capturing vein patterns in the back of the hand, palm, and wrist. It is more tolerant to changes in environmental and body condition, but it also faces the problem of disruption due to skin features such as hairs and line patterns. An initial vein pattern biometric system is implemented. The results show that all the test subjects can be correctly identified. 1 Introduction Biometrics is the science of identifying a person using physiological or behavioral features [8]. Recently, vein pattern biometrics has attracted increasing interest from both research communities [9, 1, 5, 7] and industries [3]. A Vein Pattern is the vast network of blood vessels underneath a person s skin. Anatomically, aside from surgical intervention, the shape of vascular patterns in the same part of the body is distinct from each other [6], and it is very stable over a long period of time. In addition, as the blood vessels are hidden underneath the skin and are mostly invisible to the human eye, vein patterns are much harder for intruders to copy as compared to other biometric features. The properties of uniqueness, stability and strong immunity to forgery of the vein pattern make it a potentially good biometrics which offers secure and reliable features for person identity verification. In many medical practices, X-ray and ultrasonic scanning are used to form vascular images. Whilst these methods can produce high quality images for blood vessels, it is an invasive technique as it requires injection of agents into the blood vessels. This is not acceptable for general purpose biometric applications in the real-world. Therefore, obtaining the vein pattern images in a fast and non-invasive manner is the key challenge in a vein pattern biometric system. However, no research has specifically addressed the issue of vein pattern acquisition, and there is a lack of analysis of the factors affecting the quality of the vein pattern images. This paper is thus motivated to investigate the utilization of infrared imaging technologies in this area of application. This paper focuses on two types of infrared imaging techniques: Far-Infrared (FIR) and Near-Infrared (NIR). As will be discussed in the later sections, Infrared imaging in these two regions is capable of capturing the superficial vein patterns inside the human body. It provides a contact-less, non-invasive data acquisition method and requires no injection of any agents into the blood vessels. Therefore, it is by far the best known non-invasive option to acquire vein pattern images. To fully investigate these two imaging methods, the region of study is the vein patterns in the various parts of hands. The paper is organized as follows: Sections II and III investigate in detail, both theoretically and experimentally, the utilization of far- and near- infrared imaging technologies respectively for capturing vein patterns in various parts of the hand. Following this, in Section IV, the paper reports a database built for the research. To verify the usefulness of the infrared vein pattern images for biometrics, an initial biometric system based on far-infrared vein images is proposed, implemented and evaluated in Section V. Section VI gives concluding remarks of this research.

2 2 FIR Imaging of Hand Vein Patterns Almost all objects emit infrared radiation when they are heated. The Far-Infrared (FIR) imaging technology forms an image passively using the infrared radiation emitted by the human body. 2.1 Principle of Imaging All objects radiate a continuous spectrum of frequencies. The total emissive power is described by the Stefan- Boltzmann Law given in Equation 1, where ε is the emissivity of the object and σ = watt/m 2 K 4 is Stefan s constant. The relationship between the wavelength λ and black body temperature T is formulated by Wiens Displacement Law based on Plancks energy distribution law as given in Equation 2. w = ε σ T 4. (1) λ max = /T. (2) Typically, a human body emits infrared radiation with wavelength in the range of 3 14 µm. 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 µm and 8 14 µm, 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 µm or 8 14 µm, an image showing the heat distribution of the human body can be generated. Medical researchers have observed (and this is also quite intuitive) that superficial human 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 System Setup and Image Acquisition In this research, an NEC Thermo Tracer TS7302 was used as the FIR image acquisition device. It operates in the spectral range of 8 14 µm, and has temperature and image resolution of 0.08 C and 320(H) x 240(V) respectively. The camera was mounted on a copy-stand and adjusted to a height of approximately 30cm above the experimental board of the copy-stand. The camera was connected to a workstation. During the FIR image acquisition, the subject placed one hand on the experimental board with the back of the hand facing upwards; then the FIR camera was used to capture the temperature profile of the hand. Once captured, the thermo analyzer software will convert the temperature data into gray-scale images for further analysis. 2.3 FIR Vein Image Quality Analysis Figures 2(a) and 1 show some vein pattern images captured using this FIR imaging method. From Fig. 2(a), It can be seen that the major vascular network in the back of the hand is successfully captured. However, for the palm side (Figure 1(a)) and the wrist area (Figure 1(b)), there is no observable meaningful information of the vein patterns contained in the image. The images in Figure 2(a) were captured in a normal office environment, where the ambient temperature and humidity are constant and the ambient temperature is at least 10 degrees centigrade less than the human body temperature. Figure 2(b) shows another set of images captured in a tropical outdoor environment (30 34 C and > 80% humidity). It can be seen that the ambient temperature and humidity have a negative impact on the image quality, and the vein patterns in these images are now not easily visually distinguishable. In addition, the body condition also affects the image quality. Figure 3 shows the images of the same hand at two different time instances. It is apparent that Far-Infrared imaging technology is very sensitive to external conditions. Overall, most of the FIR images have low levels of contrast, which makes it difficult to separate the veins from the background. Also, due to heat radiation, the tissue near the blood vessels has similar temperature (and hence similar gray level in the image) as the vein. This makes it difficult to locate the exact position of a vein. In addition, as the Far-Infrared imaging can only capture the major vein patterns, the information contained in these patterns is limited, which will prevent it from being used as the only biometric in any high security application. 3 NIR Imaging of Hand Vein Patterns Human eyes can only detect visible light that occupies a very narrow band (approx nm wavelength) of the entire electromagnetic spectrum. However, generally speaking, there is much more information contained in other bands of the electromagnetic spectrum reflected by the objects of interest. For human vein patterns beneath the skin, their visibility under normal visible light conditions is fairly low. However, this can be resolved by using Near- Infrared (NIR) imaging techniques. 3.1 Principle of Imaging Two special attributes of infrared radiation and human veins lead to a novel method of vein pattern imaging: (i) The incident infrared light can penetrate into the biological tissue to approximately 3mm depth, and (ii) The reduced hemoglobin in venous blood absorbs more of the incident

3 (a) FIR image of the palm (b) FIR image of the wrist Figure 1. Far-Infrared image of the palm and wrist (a) FIR images of the back of the hand in an office environment (b) FIR images of the back of the hand in an outdoor environment Figure 2. Far-Infrared image of the back of the hand infrared radiation than the surrounding tissue [1]. Therefore, by shooting an infrared light beam at the desired body part, an image can be captured using a CCD camera with an attached IR filter. In the resulting image, the vein patterns appear darker than the surrounding parts and are easily discernible. Biologically, there is a medical spectral window which extends approximately from about 700 to 900 nm, where light in this spectral window penetrates deeply into tissues, thus allowing for non-invasive investigation [2]. Therefore, typically, the wavelength of the infrared light beam coming out from a light source is selected to be within the near infrared region with wavelength around 850nm. Using this wavelength, it also avoids undesirable interference from the IR radiation (with a wavelength of 3um - 14um) emitted by the human body and the environment. 3.2 System Setup and Image Acquisition In our study, we used two LED array lamps to shine infrared light onto the hand from both sides. The infrared light emitted by the LED lamps peaks at a wavelength of 850nm. In order to form an image with this reflected infrared light from the hand, we need to use a camera whose spectral response also peaks at a wavelength of around 850nm. A Hitachi KP-F2A infrared CCD camera was selected for this purpose, as the spectral sensitivity range of this camera well covers the peak of the infrared light from the LED lamp. To eliminate the effect of visible light, an optical infrared filter was mounted in front of the camera s lens. Three infrared filters with different cutoff wavelength (720nm, 800nm, and 900nm) were experimented with, and it was concluded that the infrared filter with cutoff wavelength of 800nm produced better images. Hence, during our data collection, a Hoya RM80 filter was used. The camera was mounted on a copy-stand, and adjusted to be approximately 60cm above the board. The camera was connected to a computer to capture the images using a frame grabber. Similar to the far-infrared imaging, participants were required to place their hand at the center of the experimental board of the copy-stand. At first, they placed their hand with the back of the hand facing up at the camera. Three images were then taken of the vein patterns in the back of the hand. Then the participant flipped the hand over with palm side facing up, and three more images were taken of the vein patterns in the palm. Finally, the participant moved the wrist to the center of the board, and another three images were taken of the veins in the wrist area. 3.3 NIR Vein Image Quality Analysis Figures 4 shows vein pattern images captured by the NIR camera for the three parts of the hand: the back of the hand;

4 (a) (b) Figure 3. Far-Infrared image of the same hand at different time instances the palm; and the wrist. It can be seen from Figure 4(c), the NIR imaging technique can capture the major vein patterns in the back of the hand as effectively as the FIR imaging technique. More importantly, the NIR camera is capable of capturing images of the small veins lying in the palm and wrist areas. Unlike the image of the back of the hand, where only major veins are visible, the vein pattern in the palm is far more complex and contains much more information than the one in the back of the hand. This is important because it significantly increases the discrimination power of the vein pattern biometrics when the size of the user group is large. The NIR imaging technique is more tolerant to the external environment and the subject s medical condition and environmental situation. In our study, the quality of the images does not change significantly for both air-conditioned and outdoor environment. Also, the color of the skin does not affect the visibility of the vein patterns in the image. During the study, for the white skinned Caucasian and the tan skinned Indian, their vein patterns are all visually distinguishable in the images. However, NIR imaging of vein patterns suffers from the disadvantage that the defects on the skin surface are also visible in the image, which sometimes will corrupt the structure of the vein patterns and lead to problems when it comes to the later image processing and pattern recognition stages. Figure 4(d) shows confusion in the NIR images caused by the hair on the back of the hand. Also, in Figure 4(a), the palm prints are also mixed with the vein patterns. Whilst human beings are capable of distinguishing these defects from the vein patterns in the image, it is a demanding task to remove these defects using automatic processing. 4 Database Generation To the best of our knowledge, there is no hand vein pattern database available to the public research community. Therefore, it was necessary to construct a database for further investigation of the vein pattern biometrics. A database containing both Far-infrared and Near-infrared vein pattern images was built up for the research. To improve the representation of the database, we set up the two data acquisition stations in a public venue, and we invited volunteers (both female and male) from the public to participate in our data collection experiment. In our data, the participants are of various racial groups, representing the racial mix in Singapore. They were mainly Chinese, Indian, and Caucasian but include other races such as Thai and Vietnamese. The age group of those participants was between 18 to 60 years, and their occupations ranged from university students, professors, and technicians to manual workers such as cleaning ladies and electricians. Table 1 shows the distribution of participants against the genders and races for the NIR image database, whilst Table 2 is the age distribution for both FIR and NIR image database. The database of Far-infrared vein pattern images, currently has 30 distinct subjects. Each subject provided 9 images for the vein patterns in the back of both left and right hands, which resulted in a total of 540 Far-infrared images in the database (9 x 2 x 30). The images are in 256 level gray-scale of size 320 x 240 pixels and are stored in Bitmap (bmp) format. For the Near-infrared vein pattern images, we have currently collected images from 150 subjects. The database contains NIR images for the back of the hand (3 images), the palm (3 images), and the wrist (3 images). Images were taken for both the left and the right hand. Hence, there are 2700 images in the database (9 x2x150) for the 150 subjects. The NIR images are also in 256 level gray-scale and of size 644 x 492 pixels. They are stored as Tagged Image File Format (tiff). The database is being extended gradually, but contains sufficient data for feasibility studies and evaluations to be carried out. 5 Biometric Systems Based on Far-Infrared Hand Vein Patterns To verify the suitability of infrared imaging for hand vein pattern biometrics, we conducted an initial evaluation using the Far-Infrared image database of the vein patterns in the

5 (a) NIR image of palm (b) NIR image of the wrist (c) NIR image of back of the hand (d) NIR image of the back of the hand with hair Figure 4. NIR images of various parts of the hand Table 1. Participant Distribution Against Race Groups for NIR Images Database Chinese Indian Caucasian Others Total Female Male Total Table 2. Participant Distribution Against Age Groups for both FIR and NIR Image Databases < Total FIR NIR back of the hand. Another NIR images based biometric system will be built and tested in the near future. The proposed system in the research consists of five individual processing stages: Hand Image Acquisition, Image Enhancement, Vein Pattern Segmentation, Skeletonization and Matching, as shown in Figure 5. Unlike other vein pattern verification systems that compare the vein patterns based on a predefined set of features extracted using techniques like Multiresolution analysis [7], the proposed system recognizes the shapes of the preprocessed vein patterns by calculating their line segment Hausdorff distances (LHD) [4] given in Equation 3,4 and 5, where d θ (m l i,tl j ), d (m l i,tl j ) and d (m l i,tl j ) are the angle distance, parallel distance and perpendicular distance respectively and H l is value for the undirected LHD. d(m l i,t l j)= (W a d θ (m l i,tl j ))2 + d 2 (ml i,tl j )+d2 (ml i,tl j ) (3) h l (M l,t l )= 1 m l i M l l m l i m l i M l l m l i min t l j T l d(m l i,t l j) (4) H l (M l,t l ) = max(h l (M l,t l ),h l (T l,m l )) (5) Testing was carried out on our Far-Infrared vein pattern image database, which consists of 270 left hand images from 30 people (9 from each person). Figure 6 shows the distribution of the genuine and intruder accesses against the similarity measure H (which is the LHD between the testing subject and the template). It can be readily observed from the figure that the smaller H is, the higher the probability the vein pattern belonging to the genuine class. By choosing 9.0 to be the threshold value, the system achieves a 0% false acceptance rate (FAR) and 0% false rejection rate (FRR) for all the 270 images in both the testing set (containing 180 images) and the template training set (containing 90 images). The results of the experiment are encouraging. It shows vein pattern in the back of the hand can be used as biometric features to identify a person. However, when the user population becomes large, this vein pattern will not have sufficient discriminating power. Therefore, the vein patterns in the palm side are preferred for application with large user group. Another solution to improve the system performance is to acquire the vein patterns both in the back of the hand/wrist and the palm, and to use them as a combined feature. 6 Conclusion This paper investigates Near- and Far- Infrared imaging technologies for acquiring hand vein patterns for biometric purposes. The two imaging techniques are applied to acquire vein pattern images for various parts of the hand: the back of the hand; the palm and the wrist. The experiments show that far-infrared imaging encounters difficulties in capturing vein images in the palm and wrist. However, it can capture the large veins in the back of the hand quite

6 Image Acquisition Raw Images Image Enhancement & ROI Selection Finer Images Vein Pattern Segmentation Skeletonization Vein Pattern Shape Match Decision Data Collection Vein Pattern Extraction Template Database Figure 5. Hand vein pattern verification system model H' Geniune Intruder Access Attempts Figure 6. Distribution of genuine and intruder accesses against similarity measure H well, but it is sensitive to ambient conditions and human body condition. On the other hand, near-infrared imaging outperforms the FIR imaging and produces good quality images when capturing vein patterns in the back of the hand, palm, and wrist. It is more tolerant to the change of environment and body condition, but it also faces the problem of image corruption from skin features such hair and lines in the skin. An initial biometric system is proposed to test the vein pattern database. The experimental results all the testing subjects could be correctly identified, which demonstrates that vein pattern biometrics with infrared imaging is a potentially good biometrics. References [1] J. Cross and C. Smith. Thermographic imaging of subcutaneous vascular network of the back of the hand for biometric identification. In Proceedings of IEEE 29th International Carnahan Conference on Security Technology, pages 20 35, Sanderstead, Surrey, England, October [2] S. Fantini and M. A. Franceschini. Handbook of Optical Biomedical Diagnostics, chapter 7. SPIE Press, Bellingham, WA, [3] Fujitsu-Laboratories-Ltd. Fujitsu laboratories develops technology for world s first contactless palm vein pattern biometric authentication system. Online: March [4] Y. Gao and M. Leung. Line segment hausdorff distance on face matching. Pattern Recognition, 35: , [5] S.-K. Im, H.-M. Park, S.-W. Kim, C.-K. Chung, and H.-S. Choi. Improved vein pattern extracting algorithm and its implementation. In Digest of technical papers of International Conference on Consumer Electronics, pages 2 3, Los Angeles, US, June [6] A. Jain, R. Bolle, and S. Pankanti. Biometrics: Personal Identification In Networked Society. Kluwer Academic Publishers, Dordrecht, [7] C.-L. Lin and K.-C. Fan. Biometric verification using thermal images of palm-dorsa vein patterns. IEEE Trans. Circuits and Systems for Video Technology, 14(2): , [8] N.K.Ratha,A.W.Senior,andR.M.Bolle. Tutorialonautomated biometrics. In Proceedings of International Conference on Advances in Pattern Recognition, pages , Rio de Janeiro, Brazil, March [9] L. Wang and C. G. Leedham. A thermal hand vein pattern verification system. In S. Singh, M. Singh, C. Apte, and P. Perner, editors, Pattern Recognition and Image Analysis, volume 3687 of Lecture Notes in Computer Science, pages Springer, 2005.