diplomsko delo
Abstract
V tem diplomskem delu smo implementirali aplikacijo za odklepanje telefona z Android operacijskim sistemom s pomočjo prepoznave obraza. Na začetku smo pregledali že obstoječe metode iz področja prepoznave obrazov, nato pa smo podrobneje opisali naš postopek za prepoznavo obraza. V našem postopku smo najprej detektirali pomembna območja na obrazu, ki smo jih nato obdelali s pomočjo Gaborjevih filtrov in uniformnih lokalnih binarnih vzorcev. Dobljene vrednosti smo shranili v vektor značilk. Pri fazi prepoznave smo uporabili Pearsonovo mero različnosti za izračun razdalje, vzorce pa smo razvrščali po metodi najbližjega soseda. Prag za razvrščevalnik smo izračunali s pomočjo povprečja razlik med slikami v učni množici. Algoritem je bil testiran na desetih osebah, ki smo jih slikali skupno 85-krat. Naš algoritem daje dobre rezultate ob dobro osvetljenih slikah (dnevna svetloba – natančnost okoli 90 %), natančnost pa se zmanjša pri testiranju ob slabših osvetlitvenih pogojih (sobna luč – 77,70 % natančnost). Aplikacija se dobro obnese pri zavračanju obrazov, ki jih ne smemo prepoznati (specifičnost – 90,70 %). Rezultati so nekoliko manj uspešni pri prepoznavi obrazov, ki jih moramo prepoznati (občutljivost – 85,71 %).
Keywords
razpoznavanje obraza;Android OS;gaborjevi filtri;uniformirani lokalni binarni vzorci;
Data
Language: |
Slovenian |
Year of publishing: |
2015 |
Typology: |
2.11 - Undergraduate Thesis |
Organization: |
UM FERI - Faculty of Electrical Engineering and Computer Science |
Publisher: |
M. Hazl |
UDC: |
621.395.721.5(043.2) |
COBISS: |
19318550
|
Views: |
1047 |
Downloads: |
198 |
Average score: |
0 (0 votes) |
Metadata: |
|
Other data
Secondary language: |
English |
Secondary title: |
UNLOCKING A SMARTPHONE WITH ANDROID OPERATING SYSTEM BY USING FACE RECOGNITION OF AN USER FROM A DIGITAL PICTURE |
Secondary abstract: |
In this thesis we implemented an application which unlocks a phone with the Android operating system by face recognition. Firstly, we reviewed existing methods from the research area of face recognition and then we described our procedure for face recognition in more detail. In our algorithm we started with detecting important areas of the face which we then processed with the help of Gabor filters and uniform local binary patterns. The results we obtained were then saved to feature vector. In the recognition stage we used Pearson's dissimilarity measure to calculate the distance. For sample classification we used the nearest neighbour method. We calculated the classification threshold from the mean value of the distances between pictures in our training set. The algorithm was tested on ten different people of which we took 85 pictures. Our algorithm produces good results with pictures that are well illuminated (daylight – accuracy approximately 90 %) but the performance is reduced if the pictures are taken in poor lighting conditions (indoor lighting – 77.78 % accuracy). Application is good at rejecting faces that should not be recognized (specificity – 90.70 %). The results are slightly worse when recognizing the faces that should be recognized (sensitivity – 85.71 %). |
Secondary keywords: |
face recognition;Android OS;Gabor filters;uniform local binary pattern; |
URN: |
URN:SI:UM: |
Type (COBISS): |
Bachelor thesis/paper |
Thesis comment: |
Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko, Računalništvo in informacijske tehnologije |
Pages: |
VII, 38 f. |
ID: |
8887448 |