Research On Palmprint Identification And Implement On DSP

Palmprint recognition technology is biometric identification in the field of emerging technologies, it is defined as the effective manpower handheld information to identify palmprint image which is from the same hand or different hand. Palmprint recognition system is currently in a variety of biometric systems at a very small amount of the very few civilian palmprint recognition system, in the domestic civilian market palmprint recognition system is almost blank. Then the development of small, cheap, short development cycle, the maintenance of convenience, to meet the medium / low-level security requirements of the palmprint identification system in order to promote the civil and wider application of the field will be a very meaningful study.There are a lot of Palmprint recognition core algorithm, based on statistical methods to identify the characteristics of the main FFT transform, wavelet transform, K_L transformation. In this paper, FFT transform the light of the principle of simple, easy to achieve in order to transform FFT-based palmprint recognition method, the drawback is that the recognition rate of FFT transform based is much lower than the recognition rate of wavelet transform based.Palmprint identification of the core ideas of this article is based on the radius of the palm of the hand for a rough match, FFT transform spectrum image in the 8 direction of the energy adder values to a detailed match. This paper described in detail the positioning rotating, spliting and enhance pre-treatment process of palmprint images,and achieved Palmprint matching process.In the pre-processing stage, according to proposed principle of using the morphological image of palm prints ideological positioning to achieve its algorithm, and description of its advantages and disadvantages. According to proposed based on the " translational shear" the principle of rotation of the image method, the paper did not do a simplified one-dimensional linear interpolation to deal with the simplified algorithm has done a detailed description of the advantages and disadvantages, and the angle of rotation to avoid decomposition of the effective methods in the wide-angle image rotation when the issue of missing pixels This article uses proposed modified template based on the 8 direction Sobel operator to do the image edge detection, edge effect has been very good, in view of improving to do Sobel edge detection operator on the edge of the inhibitory effect of noise prominent characteristics of the edge sharpening distinctive features, this paper presents the use of Sobel operator to do to improve the image enhancement method...Stage in the match, as the use of image segmentation Laplacian image enhancement to do directly after the FFT transform of the spectrum characteristic of its poor differentiation, so in the extraction of the spectrum characteristic of the former, namely the use of a combination of classical bimodal threshold Sobel algorithm value extracted palmprint 2 value image to doing FFT transform, Sobel algorithm to improve the combined use of single-threshold ridge extracted palmprint 2 value image to doing FFT transform, Sobel image enhancement to improve image processing directly after the FFT transform, These three different pre-processing method for pre-processing algorithm of the three match test results, Sobel classic combination of dual-threshold algorithm for the system to deal with the lowest recognition rate, error rate to accept the lowest, the highest rate of error to refuse to improve the Sobel Algorithm single-threshold system to deal with the highest recognition rate, error rate to accept the highest, the lowest error to refuse to improve the image enhancement Sobel operator approach in the recognition rate between the two. Practical applications, it is necessary to identify three methods based on the characteristics of the system application environment combining the three to choose the most suitable method.Hardware platform is the DM642 DSP development board, in the code is written in assembly line cycle control, such as register allocation optimization done, procedures to shorten the running time from 7.9s to 1.8s, effective optimization to meet real-time requirements.