Научная визуализация

Scientific Visualization

Электронный журнал открытого доступа

 Национальный Исследовательский Ядерный Университет "МИФИ"

      ISSN 2079-3537      

 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                                                                                             





Научная визуализация, 2018, том 10, номер 5, страницы 1 - 12, DOI: 10.26583/sv.10.5.01

The Performance Improvement on 4-FB Face
using Random Forest Classifier

Авторы: B. Nassih1, A. Amine2, M. Ngadi3, D. Naji4, N. Hmina5

LGS, National School of Applied Sciences, Ibn Tofail University, B.P. 241, university campus, Kenitra, Morocco

1 ORCID: 0000-0002-2389-4380, nassih.bouchra@univ-ibntofail.ac.ma

2 ORCID: 0000-0002-6806-3513

3 ORCID: 0000-0002-4632-0531

4 ORCID: 0000-0002-3279-0241

5 ORCID: 0000-0002-7739-4875

 

Аннотация

The growing demand in the field of security leads to the development of interesting approaches in face classification. For this reason, we bring a new method based on the extracted features of the four Frequency Blocks (4-FB) and Random Forest (RF) to classify faces and non faces, thus, we have used the fusion of three extracted features based on Discrete Cosines Transform (DCT), Uniform Local Binary Pattern (ULBP), and Histogram of Selected Regions (HSR), using RF classifier then, compared to other classifiers such as ID3, C4.5, K Nearest Neighbor (KNN), and Neural Network (NN). Firstly, we have used each descriptor separately, secondly, we have combined the three descriptors in pairs and in trinomial, to evaluate the performance of the proposed method. Our new approach is to apply RF classifier only on local features, where, exhibits pertinent information about the face image as eyes, mouth, and nose which displays by four Frequency Blocks (4-FB). The performance of our proposed method was evaluated on our created database named BOSS. Moreover, RF classifier combined with DCT and ULBP provided better results with 4-FB compared to other used methods which exceeds 99%.

 

Ключевые слова: Face Classification, HSR, DCT, ULBP, RF, ID3, C4.5, KNN, NN, Feature Extraction, Feature Fusion.