Description

ABSTRACT

The recent advancements in the field of image processing have led to giving the importance of kinship verification. In this paper, we propose a methodology based on Deep Neural Networks (DNN) and Support Vector Machine (SVM) classifier for the identification of human faces in the images. We examined the best parameters for every feature extraction technique such as Grey-Level Co-Occurrence Matrix (GLCM), Completed Joint Scale Local Binary Pattern (CJLBP), Alexnet, Resnet on datasets. The method is made up of two basic stages which are; (1) Feature Extraction (2) Classification. In the proposed method we adopted Alexnet for the process of feature extraction, and the classifier used is support vector machine (SVM). The results obtained using the proposed approach give better results in comparison to many other approaches that were used previously.

INTRODUCTION

In recent years the field of image processing has developed due to which face verification has also emerged. Previously, the parent-child verification was done through biometrics, Various applications of facial recognition include images observations, analysis of social sites based on images, categorizing the photographs, checking likeness among humans, searching databases for parents, and searching for a missing person or kid. In this paper, the approach for facial verification through images is based on the Grey-Level Co-Occurrence Matrix (GLCM) descriptor and Alexnet Deep Neural network model. the First step was to extract the features with descriptors such as GLCM, CJLBP and also through Alexnet, Resnet and Googlenet Deep Neural Network model. The obtained features were then fed to Pattern Recognition Tool but later on SVM classifier was used for final results. The process of kinship verification is to check whether the two people are genetically identical or they have any family relationship between them. Our system is based on two stages; (1) Feature Extraction (2) Classification and Decision (kin or non-kin).

PROPOSED SYSTEM

In this paper, we propose a methodology based on Deep Neural Networks (DNN) and Support Vector Machine (SVM) classifier for the identification of human faces in the images. We examined the best parameters for every feature extraction technique such as Grey-Level Co-Occurrence Matrix (GLCM), Completed Joint Scale Local Binary Pattern (CJLBP), Alexnet, Resnet on datasets

BLOCK DIAGRAM

CONCLUSION

In this work we proposed a new approach using the deep neural network, we used many models like Google net, Resnet-50, Inceptionv3, but the best results were received by the Alexnet deep neural network. The results approved effectiveness as compared to the other techniques that were previously used.