Monica* et al. (IJITR) INTERNATIONAL JOURNAL OF INNOVATIVE TECHNOLOGY AND RESEARCH Volume No.4, Issue No.3, April – May 2016, 3025 – 3032. 2320 –5547 @ 2013-2016 http://www.ijitr.com All rights Reserved. Page | 3025 A Comparative Analysis of Object Recognition System Using SIFT, SURF and FAST Algorithms MONICA P.G Student Department of Electronics & Communication Engineering SKITM Engineering College Bahadurgarh, Haryana, India RAVIKANT KAUSHIK Asst. Professor Department of Electronics & Communication Engineering SKITM Engineering College Bahadurgarh, Haryana, India RITU JANGRA Asst. Professor Department of Electronics & Communication Engineering HMRITM Engineering College Delhi, India Abstract— Object recognition has always been an area of interest for various researchers since decades. In this paper an attempt has been made to give a comparison between various techniques of object recognition mainly feature based approaches. In this paper an overview of the Famous and impressive technique by David Lowe, which is Scale Invariant Feature Transform (SIFT) has been given. Another very important technique called Speeded-Up Robust Feature Transform (SURF) has been used to conclude with certain interesting results. FAST is the third technique which has also been discussed in this paper. SIFT, SURF and FAST algorithms has been implemented on COIL dataset and a comparative analysis of these techniques has been given. The algorithms has been evaluated on two parameters i.e., number of features extracted and the time of execution. It has been seen that SIFT has detected more number of features as compared to SIFT and FAST. But the times of execution taken by SURF is comparatively less than SIFT and SURF. Keywords - Object Recognition; Descriptors; Feature Extraction; SIFT; SURF; FAST Methods I. INTRODUCTION Object recognition is a process of distinguish a particular object in an image or video. Basically object recognition algorithms based on matching, pattern recognition or feature-based techniques. Object recognition basically involves two processes: Identification and localization. It is useful in video stabilization, cell counting in bio-imaging and automated vehicle parking systems. Recently a lot of progress has been made in object categorization from images [1]. Object recognition is the subfield of computer vision which deals with recognizing the 3 D objects from image data. It also approximates the position and orientation of the recognized objects in the 3D world. Basically feature extraction is an important factor while object is required to be recognized. Feature extraction is a type in which dimension is reduced that efficiently represents interesting parts of an image as a compact feature vector. This term is also very important in image processing. This approach is useful when size of image is large and a reduced feature representation is required to complete tasks completely such as image matching and retrieval. In this paper we introduce three types of methods SIFT SURF and FAST to extract features of an image. Different results we will obtain in this approach. Object recognition in the field of computer vision describes the task to find and identifying objects in an image or video sequence form. Humans recognize a multitude of objects in images with little effort even when they are translated or rotated. Dickinson et al [2] presents a system of object recognition which involves extraction of features and then grouping them. After this he performed object hypothesis generation and finally an object verification stage. Then, Shapiro and Stockman [3] gave a typical object recognition system which was composed of: Low-level image detection and localization of objects but also the recognition and understanding of the object/stimulus in the scene. Object recognition involves three types of approaches: (i) View-based (ii) Feature- based (iii) Model-based View-based methods learn a model of object's appearance in two-dimensional image under different shapes and illumination condition. A number of view-based methods have been developed to recognize three dimensional objects. A full view of three dimensional structures can be drawn if enough two dimensional views of the object are provided. Feature-based classifies images of object under variation by rotation, noise and scaling. It is robustly and efficiently recognizes a large database of objects. It is achieved by calculating a number of features and combining them into a feature vector. brought to you by CORE View metadata, citation and similar papers at core.ac.uk provided by International Journal of Innovative Technology and Research (IJITR)