Research On Fingerprint Recognition Algorithm Based On The Least Absolute Deviation

With the popularization of computers, security has become the first consideration in modern computer systems. The confirmation of web users’ personal identity and permissions is especially important. The matter of user authentication is the common concern in current information area, which has a universal application in the fields of economy, social security and cultural life. Traditional identity recognition techniques aim to identify holdings or specified knowledge in nature, but not validate the status. Once others obtain the specific holdings or knowledge, he/she will have the same rights as the owner. Lost and stolen specific objects or knowledge has caused great losses to the society and individuals. The traditional identification techniques have been unable to meet the security demands of people in the information age. Due to such unique advantages as stability, uniqueness, versatility, non-repeatability, easy collection and so on, the fingerprint has been widely used in the field of identity recognition. Fingerprint identification technique is used as a kind of the earliest, the most widely used one, although there have been many fingerprint system products, but still cannot meet the performance requirements of some specific applications, such as on-line real-time fingerprint identification system, stained and defected fingerprint identification system, etc.This article directs against fingerprint image’s characteristics, designs an algorithm framework of fingerprint identification based on least absolute deviation, and carries out simulation experiments. The main work is as follows:Ⅱ. This paper designed an adaptive normalization algorithm for fingerprint images.Ⅲ. This paper improve a Gabor filtering algorithm based on local image blocks. An image is divided the image into N×N image blocks, and calculates their direction fields respectively. Then Gabor filter with corresponding to the main angles of an block is implemented on the block. This algorithm improves the execution efficiency.Ⅲ. The finger image is distinguished based on the least absolute deviation. Popular fingerprint match algorithms rely on the quality of the image, and extracting minutiae from the bad-quality image will cause many false minutiae to affect fingerprint identification effect. Besides, this kind of algorithm is difficult to identify tainted and defected fingerprints, especially center-missing fingerprints. This paper adopts the robust least absolute deviation fitting method to extract minutia features, and form feature coding. Take advantage of1-norm calculation to input the whole deviation of fingerprint feature coding and template feature coding, and the two fingerprint images can be defined to come from the same finger so long as the whole deviation value is smaller than the preset threshold value.