Hand Vein Recognition Based On Secondary Identification And Local Information And Feature Fusion

Vein recognition is a near-infrared biometric recognition technology, which utilizes the characteristic that the hemoglobin in the blood can absorb near-infrared light. This technology captures the personal vein distribution image first by an infrared light-sensitive device, and then extracts and matches vein features using the specific vein recognition algorithm, which realizes the personal identity authentication. In general, vein recognition includes finger vein recognition, palm vein recognition and palm-dorsal vein recognition. Compared to other biometric technologies, such as face recognition, finger print recognition and so on, vein recognition uses living biological features, which is difficult to be copied and forged. So it is more secure and reliable. Besides, vein recognition is not subject to external environment conditions, such as rough skin, and it is a non-invasive and non-contact capture progress, which proves more friendly and hygienic. In recent years, vein recognition has been paid more and more attention and become one of the mainstream biometrics identification methods.The research of this paper covers the entire progress of hand vein recognition. It includes vein image acquisition, image preprocessing, and the specific area extraction and so on. This paper proposes new vein recognition methods, which includes combining vein spatial distribution and vein coding in frequency domain, utilizing local information, using multi-feature fusion and so on. Comparing with many other state-of-the-art vein recognition algorithms, the proposed methods have achieved a better recognition result. The main work and contributions of this paper is as follows:1 The hand vein recognition method based on image segmentation with secondary identification has been proposed. After the first recognition result is gotten by encoding, extracting and matching features in frequency domain, the method obtains the structure of vein by segmentation, puts forward a new index to evaluate the similarity of two vein structures, and second recognizes the hand vein according to the index. The proposed method with secondary identification applies the image segment algorithm to the vein recognition, which combines vein spatial distribution and coding in frequency domain. Thus it can correct the first preliminary identification result, and improves the recognition accuracy.2 The hand vein recognition method using local block pattern has been proposed. Local features represents discriminative texture information. To take full advantage of these features, the method adopts the multi-level partition strategy. Firstly, the vein image is divided into blocks, and then each block is divided into sub-blocks further. Histogram statistics is limited in a scope of sub-block, which can maximize the local differences and reduce the noise weight. Feature representation is limited in a scope of block, which can avoid "too high dimension" and "too small sample" problems. In the progress of feature extraction, Fisher linear discriminant are adopted to reduce the feature dimension. The good balance between the global and the local features significantly improves the accuracy of vein recognition.3 The hand vein recognition method based on the fusion of Gabor amplitude feature, phase feature and other features has been proposed. The Gabor ordinal measure uses eight encoding masks, extracts and obtains the amplitude feature, the phase feature, the real part feature and the imaginary part feature of vein image. And then based on the local block pattern above, the method gets the feature histogram of each block by concatenating four types of Gabor feature histograms, and calculates the similarity of two vein images. The proposed hand vein recognition method with multi-feature fusion utilizes the differences of Gabor each feature, enhances the discriminative power and the anti-noise ability of the feature descriptor, and achieves a higher accuracy.Besides, for assisting hand vein recognition in identity certification, the author is also engaged in the research of face and facial expression recognition with other researchers, and has obtained a gradual achievement.