New features in extraction techniques of low-quality fingerprints

العناوين الأخرى

تقنية جديدة لاستخراج الميزات من صور البصمة منخفضة الجودة

مقدم أطروحة جامعية

al-Shamailh, Ahmad Jamil Muhammad

مشرف أطروحة جامعية

al-Kasasibah, Muhammad

الجامعة

جامعة مؤتة

الكلية

كلية تكنولوجيا المعلومات

دولة الجامعة

الأردن

الدرجة العلمية

ماجستير

تاريخ الدرجة العلمية

2016

الملخص الإنجليزي

Identifying individuals through fingerprints depends on two factors: the technique that have used to extract features and the technique that have been used in the comparison phase.

Although there are many algorithms have used, it primarily depends on minutiae.

Feature extraction is a main stage in fingerprint recognition, because of its impact on the performance of systems.

Two major contributions have achieved in this thesis: first, a newfeature extraction method for digital images of fingerprints hasproposed, which depends on the edges of the component that forms a fingerprint used to identify individuals;second, the creation of a new lowquality database,theSmartphone Fingerprint Database (SPF).A mobile phone camera has used to capture the fingerprints of 100 different people;5 samples for each person totalling500 fingerprint images.

Low-quality images haveconverted to grayscaleimages, thenaedge detection methodhasused to find the gradient edges of the fingerprint components.

The image hasdivided into 5*5-sized blocks.

Theiron filing hasthen thrown in the center of each block and attracted the maximum magnetic power; any values less than a specific thresholdhave removed.

The rest attracted magnetic power (magnitude) hasconsidered as new features, which saved as vectors for later comparison.

After the feature extraction two methods have used for matching between vectors.First,Dynamic Time Warping (DTW) has used to compute distance, then the system performance has evaluated based on a K Nearest Neighbor KNN classifier.

The secondmethod represented features using a histogram of the probability density function (PDF) to prepare data for machine learning classifiers.

The results have proved that relying on the edges as features can be adopted as an effective method for fingerprint systems.The proposed model has achieved satisfying results depending on different threshold values that were applied.The results have showed the importance of the proposed method which achieved a 94.2% accuracy rate when using a 35 threshold.

This studyalso has proved empirically that using whole features from all components of the image resulted in better accuracy results than excluding the features on the boundary of the edge.

This study has showed that using theoriginal features achieved better results than using preprocessed data.It also showedthat the results were negatively affected by using the Principal Component Analysis (PCA) as a preprocessing technique.

التخصصات الرئيسية

تكنولوجيا المعلومات وعلم الحاسوب

عدد الصفحات

53

قائمة المحتويات

• APA
• MLA
• AMA

لغة النص

الإنجليزية

نوع البيانات

رسائل جامعية

رقم السجل

BIM-749292