Implementing the Data Mining Approaches to Classify the Images with Visual Words
Annaluri Sreenivasa Rao1, Attili Venkata Ramana2, S. Ramakrishna3
1Dr. Annaluri Sreenivasa Rao, Department of Information Technology, Vallurupalli Nageswara Rao Vignana Jyothi Institute of Engineering & Technology, Hyderabad (Telangana), India.
2Dr. Attili Venkata Ramana, Associate Professor, Department of ECM, Sreenidhi Institute of Science and Technology, Hyderabad, (Telangana), India.
3Dr. S. Ramakrishna, Professor, Department of Computer Science, SVUCM & CS, SV University, Tirupati (A.P), India.
Manuscript received on 30 March 2019 | Revised Manuscript received on 09 April 2019 | Manuscript Published on 27 April 2019 | PP: 901-909 | Volume-7 Issue-6S2 April 2019 | Retrieval Number: F11180476S219/2019©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: In the recent times, image classification got a significant importance and is going through wide range of challenges. In this work, a unique system is built to reveal the information of the image based on its category and relevant labels associated with it. To accomplish this task different data mining techniques are used for image classifications using Bag of Visual Words (BoVW) feature extraction algorithm. This algorithm is constructed with the help of grey level features and along with some of the colour features. Grey level features include speed up robust features (SURF), maximum stable external regions (MESR) and improved colour coherence vector (ICCV). Apart from that different techniques involved in data mining are being used includes neural networks (NN), decision trees (DT), Bayesian networks (BN), discriminant analysis (DA) and K-nearest neighbour (KNN). These techniques are used to evaluate the datasets included with Corel-1000 and COIL-100. BN and DA outperformed other methods as they are achieved a specificity of 99.9%, sensitivity of 99.5% and accuracy of 99.4% for Coral-1000. Whereas the same experimented results reported to be 100%, 98.5% and 98.9% respectively for COIL-100.
Keywords: Data mining, Machine Learning, Image Classification, NN, KNN, BN, DA.
Scope of the Article: Data Mining