Research On Algorithm For Human Dorsal Hand Vein Recognition

The development of internet and information promote the advancement of modern society, and digital, information, network as the characteristic of the development in this society has proposed the band-new demand to security. In this environment, traditional position recognition displays much leak that will not be solved in any way, and today, people expect one kind of low cost, safer, more convenient and reliable ways. This technology is"Biometrics Recognition", because our own body is the safest, most reliable key and password.The Biometrics Recognition is the technology to distinguish peoples'position automatically using the individual's physiology characteristic (such as fingerprint, iris, vein etc.) and behavior(for example writing signature, sound etc.) by the way of image processing and pattern recognition. The characteristic includes the human body living things mainly: Step state, waits sound, fingerprint, palm print, iris, vein, DNA, signature etc. The core of Biometrics Recognition is how to gain these living things characteristics and change them to the digital data for memory, at the same time with matching algorithm peoples'position automatically is done.In different biological characteristic, the dorsal hand vein is the most reliable and most lasting one. The dorsal hand vein recognition has more outstanding natures comparing with other living things characteristics: uniqueness, the stability, non- contact gathering, not touching, against false, low cost, uninfluenced by injury to epidermis and so on. Not touching gathering is the evitable trend of developing status recognition research and application. The dorsal hand vein recognition has more precise result comparing with face and sound. Vein status recognition technology gradually obtains the academic circles and enterprise's value. Because the dorsal hand vein in the imagery process is influenced by equipment, the gathering time, hand thickness, the luminous intensity, and rotation and so on, which enable this technology to become a new challenge research topic in the pattern recognition and the machine vision.This article analyzes dorsal hand vein image characteristic, and has thorough research on the essential technology and the core algorithm of the dorsal hand vein recognition system. Specially, in hand vein image gathering, image processing, feature extraction, matching we proposed a series of new or improved algorithm, and obtains some ideal recognition results. This article main research results include:1. This article has designed a set of simple dorsal hand vein image gathering experiment equipment, and this equipment can completely satisfy the hand vein recognition system to gain the request of image gathering, and has the product foundation. The image gathering equipment photo source is a near-infrared light emitter diode array, and the launch wave length is 810mm. When light emitter diode array launch near far infrared light penetrates dorsal hand, at the same time the light intensity is received by artificial adjust, and the CCD camera gathering image with a neutral density light filter for removing background intensity. On the basis of hand characteristic, we designed a special platform to fix the hand, in this platform being equipped with a rectangular frame. When the hand puts on the platform and grips one side of the rectangular frame, at this time the dorsal hand presents the condition of a fist while the thumb is being hidden by other all fingers. Treats the hand back putting steadily when the CCD source automatically opens, this time we can carry on gathering the vein image. 2. Image processing. First, as a result of the image exposes insufficiently, the image may be dark, the image histogram gathers in some quite small value regions; similarly, the image exposes sufficiently, the picture can be bright, the image histogram gathers in some quite big value regions. Carrying on the preprocessing using the histogram balanced method, the heart thought is to turn the regions gray centralized to the ones gray uniform distribution, and the contrast is increased. Secondly because at different time the vein images gathering from an identical hand can have some rotation and translation, therefore before features extracted, we should first carry on the standardizing operation of image's position as well as the size. Finally, the vein images may be filled with the burr, and then they will be able to have the obvious disturbance. If we use the mean value filter method to eliminate the burr, it will cause the boundary to be smoothed. Unsharp masking's main goal is to highlight the detail of vein image or the fuzzy details. This kind of fuzzy production is not because of operation, but is the influence at the process of image gathering.3. Features extraction. The dorsal hand vein image may be approximately regarded as some limited strips which are different length, different direction. In the view of the computation, if carry on some kind of transformation to the image data, we will obtain most features that can reflect the vein's essential characteristic. It is one good method. While the vein image has very strong straight linear characteristic, the hand vein recognition is suitable to ridgelet theory and FRIT method. The experimental results indicated that, using FRIT method to extract he hand vein image features of the straight line, and its effect must greatly surpass other similar methods especially in the noise serious situation. So we put up an AFRIT method to extract the features of vein image, and then we obtain multi-scale ridgelet features in different scales. Secondly, this paper proposed a FRAT method to extract the features of vein image, and furthermore we extract texture direction method from usual images. Traditional Hough transformation is one effective method for extracting lines. This algorithm merit has strong ability to resist noises, and it can exact straight lines in a lower noise ratio situation. But its shortcoming is big calculation and reservation. This method has limited the algorithm processing speed in the certain degree. But because the image's dimension is always quite big, it can not fundamentally solve the speed problem of algorithm computation. While this article's method can accurately extract the texture direction and intercept, and the algorithm running rate is fast. But the shortcoming is not able to detect the line segments. In other words, it cannot provide the beginning point, the end position and the length. Experiment proved that this algorithm is extremely effective.4. Matching. The dorsal hand vein's structure is similar to "the tree" shape, and the texture has strong direction. Therefore the job of the vein pattern's matching should analyze from different angle and different surfaces carefully. This paper we proposed two matching method, that is multi-scale ridgelet feature matching and texture direction feature matching according to their features extraction method. We carried on the comparison through the experiments and the data between two methods, and produced the systematic recognition speed analysis and the recognition precision under the different match algorithms. In order to enhance the recognition precision of the system, we designed many kinds of policy-making fusion strategies to synthesize above two features recognition capability. The experimental result indicated that, the suitable policy-making fusion strategy may extremely enhance the recognition precision of the system.5. Result analysis. We match each test sample and each registered sample, then fetch different threshold value, thus get ROC curve. When FRR = 1.5%, the method of mean strategy FAR = 0.11%; MRF and texture direction feature are respectively0.25%and0.82%. The MTER of MRF and texture direction feature are0.78%and1.29%; The MTER of mean strategy is 0.62%. We can see the mean strategy is rather than MRF and texture direction feature.The dorsal hand vein recognition taking as a new emerging member in the area of biometrics recognition technology, has received the wide spread notice by its rich and stable characteristic. We believe it has the prospects for development. The dorsal hand vein recognition technology was most early proposed in 20 century's ends, because algorithm researches are also insufficiently, and has not walked to the large-scale application stage. But 21st century today, the computer hardware development ability already today is no longer as we have been with the computer profession's violent development. The product performance index is more and more outstanding, on the other hand, the price also unceasingly drops, and mathematics and the computer science also unceasingly develop. All these have provided the huge supports to the development and the application of the biometrics recognition technology. We believed the step of biometrics recognition technology will transform from the theory to the actual commodity more and more quickly. The production performance can be more and more reliable, and the cost also can unceasingly reduce. Among this, because hand vein own many superior characteristics, as well as the dorsal hand vein recognition algorithm unceasing development and mature, the dorsal hand vein recognition technology will certainly become a remarkable model in the biometrics recognition technology.