To search, Click below search items.

 

All Published Papers Search Service

Title

Object Recognition using Template Matching and Pre-trained convolutional neural network

Author

Qaisar Abbas†

Citation

Vol. 20  No. 8  pp. 69-79

Abstract

======= DOI: 10.22937/IJCSNS.2020.20.08.6 ======= Template Based Object Recognition (TBOR) is very active research area in different fields for finding object in a video or image. Nowadays, it is very difficult for TBOR-system to classify multiple-objects due to use of conventional machine learning algorithms and higher computational complexity for manually tuned features. According to literature review, it was noticed that most of existing TBOR system were unable to focus on multiple objects. To overcome these problems, a new TBOR approach is developed in this paper for the recognition of multiple objects by combining both techniques such as template-based matching and a pre-trained convolutional neural network (CNN) model. In the proposed TBOR system, object images are first projected onto features space known as template space that best encodes the variation among known object of templates. The template space is then defined by Eigen faces, which are the eigenvectors of set of objects. Afterwards, principal component analysis (PCA) method is applied to find the approximate aspects of objects, which are important for identification. At last, object recognition step is performed by combing template based PCA and pre-train CNN methods. A template-based matching technique was fully automatically implemented in this study through PCA analysis to initially recognize the object using correlation and phase angle methods. The recognition results are further enhanced by pre-train CNN model. Experimental results indicate that the proposed system is outperformed compared to state-of-the-art template-matching algorithms in terms of accuracy.

Keywords

Computer vision, object recognition, template matching, deep learning, convolutional neural network, principle Component Analysis

URL

http://paper.ijcsns.org/07_book/202008/20200806.pdf