Support Vector based Indoor Scene Classification Technique using Different Features

Abstract

In today’s World machines have started ruling the human beings as the machines are performing almost every task what human beings are capable of doing. Scene classification is one such term which is gaining importance where machines imitate the behaviour of a human being. Scene classification could be performed either on indoor scenes or on outdoor scenes using various techniques of feature extraction and out of these two categories indoor scene classification is considered more challenging. The indoor scene classification techniques have the problem of poor accuracy. In this work, indoor scene classification is addressed for improving the accuracy using the integration of SIFT (Scale Invariant Feature Transform), SURF (Speeded-Up Robust Feature) & Tamura features for the purpose of feature extraction and then using SVM (Support Vector Machine) for feature matching. The results of the algorithm are evaluated using MIT-Indoor dataset and the experimental results show that the proposed technique outperforms various existing techniques for indoor scene classification.