|
With the rapidly development of science and technology, traditional personal identification methods which are token-based and knowledge-based have been unable to meet the complicated needs of the real-time identification. Therefore, biometric identification technology with advantages of (1) inherent, (2) difficult to steal, (3) can not easily be forged, etc. have been increasingly researching and using in recent years. Palmprint as one superior characteristics of biometrics, has the following advantages: better collectability, abundant information, suitable for classification, obvious and stable. Palmprint identification system is user-friend, convenient, fast, effective, low-cost and extendable. Therefore, palmprint identification is concerned by researchers in last decade.Palmprint image has rich features, which includes geometry features such as the length, the width and the area of palm. Some parts of the palm such as three main principal line, wrinkles, delta points and minutiae also contain distictive features. Based on these characteristics, there are a lot of identification algorithms. Palmprint is a large area of soft skin, this can produce different nonlinear deformation in different sampling image, and make the matching based on local features (such as point features, line features)become difficult. The methods based on global statistic features are less impacted by nonlinear deformation than the methods based on local features, but it loss a lot of local feature in feature extraction stage which directly impact on the representation of features and make the identification between palmprint images of similar global features is so difficult. Therefore, finding a method which can represent both local features and global statistic features is becoming an important development trend.Following aspects are studied in this paper:(1) We proposed a new feature extraction method which based on local interesting points. This method used DoG (Difference of Gaussian) detector and SIFT (Scale Invariant Feature Transform) descriptor to extract and to describe palmprint features repectively. It is capable of extracting local and global features integratlly and effectively. The extracted features are stable, distinctive, concise and robust.(2) According to palmprint image’s deformation characteristic, this paper improved the traditional matching method based on LIPs. This method effectively reduced the complexity of algorithm.(3) By analyzing the characteristics of palmprint’s nonlinear deformation, we proposed a new palmprint identification method based on elastic model. This method reduced the affect of nonlinear deformation by the adaptive capacity of elastic model in feature matching stage, and improved the accuracy of matching.(4) This paper used a hierarchical scheme to build decision tree which optimized the performance of algorithm in decision stage and the execute time of decision-making, improved the accuracy of matching,. Accompanying with the rapid developing of information tecnology, nearly all parts of hunman life are related with personal identification. Biomatics indentification, as one of the most secure identification technologies, is attracting more and more research attentions. In recent years, biometrics is developing from the research stage to the application stage in global. Several biometrics approaches, including iris, fingerprint, face, hand geometry and so on, have been proposed. But none of them is outperfomed. Therefore, different situations require different biometrics.Palmprint identification, as a comparably new biometrics, is a popular research topic. This paper is a survey of palmprint-based biometricss indentification, which gives the comprehensive introductions and discussions of this tecnology including sampling, preprocessing, feature extraction, matching and decison.
|