Object detection using point feature matching based on SURF algorithm

Abstract

Object detection addresses detecting instances of objects of a particular category in digitized images and videos. Every class containing the object is associated with special features that assist in classifying the class to which it belongs. The algorithm used can detect local features and extract them for the concerned object, known as descriptor points [1]. Then compares those extracted features or descriptors with the presumed features of the original image. Matching process among the original image and the target image yields similar features, based on which a decision is made. The mentioned algorithm is called Speeded up Robust Features (SURF) algorithm. The algorithm is based on Scale-invariant feature transform (SIFT) algorithm. The implementation procedure of SURF algorithm along with experiment and its results are stated in the paper. Also, the accuracy of detection of objects using point feature matching methodology has been calculated by means of sensitivity and specificity parameters.