The Research On High Resolution Fingerprint Pore Extraction Model

With the rapid development of society, there is an increasing concern about security, privacy and identity fraud. Consequently, automatic fingerprint recognition systems(AFRS) comes into being, and are widely used in various fields, such as forensics, fingerprint attendance system, and so on. The popularity of the application also gradually exposed the weakness of the traditional AFRS: the existing anti-counterfeiting technology capacity is low, so the cheap set of fingerprint has brought great threat to user’s information and property safety; the precision of the system can not meet the needs of high-end applications. In order to solve the problems, we need to introduce new features into the existing technology to enhance the security of the system. Statistical analysis has shown that fingerprint pores play important roles in providing quantitative as well as qualitative data supporting more accurate and robust fingerprint recognition. At the same time,with the continuous improvement of the technology of fingerprint sensors, high-resolution fingerprint gradually came into people’s vision, providing the possibility of the pore features extraction.The extraction of fingerprint pores is a critical step, and how to design an accurate algorithm of features extraction is becoming the emphasis and hot spot in the investigative field of fingerprint. Most of the existing pore extraction methods detect pores by using a static isotropic pore model; however, their detection accuracy is not satisfactory due to the limited approximation capability of static isotropic models to various types of pores. An in-depth study was carried out in this thesis aiming at extracting high-resolution fingerprint pores feature. The main contribution of this thesis includes:A various kinds of noises are inevitably introduced when collecting fingerprint pictures, so it is necessary to carry out the fingerprint preprocessing. In this study, we analyzed various preprocessing functions including segmentation, normalization, enhancement and binarization, then make corresponding improvements according to different details which can effectively improve the quality of the fingerprint image.In the process of pores extraction, we firstly proposed a method which is optimizing the extraction algorithm based on Do G model, then compared it with the pore models based filters and fused them with a appropriate strategy to improve the accurate of extraction. In order to reduce the error rate, we select the size, contrast, filtering responses, gray value and distance as the coordinate pores’ features to construct a pores select model to remove more pseudo pores.Then, this study proposed a method to extract pores based on mathematical morphology processing. We use the multi-scale morphological transformation to get a binary ridge map, this module also provides significant information about pore locations. After this procedure, we extract the close pores based on the model using the area of connected components and the features of pores; For the open pores, we binarize and thin the image to extract and remove the pseudo pores by the situation of the bifurcation points. Thereby, we got the collection of the pores.Finally, experiments based on the methods proposed in this paper on a high resolution(1200dpi) fingerprint database were performed. The experimental results indicate that the methods outperform well-known state-of-the-art methods and can extract pores more accurately.