ROTATION AND SCALE INVARIANT OBJECT DETECTION

Abstract

Object Detection is the significant research area in the field of Computer Vision and Image Processing. Humans possess ability to quickly identify things just by looking at them. As humans, machines do not possess this intelligence for identifying objects in the scene. Object detection refers to the process of examining the images to determine the presence and location of the object. For detecting the object, machine need to possess the prior knowledge about the object i.e. model object. The task of detecting object in arbitrary environment is challenging for variety of reasons. The occurrence of object in the image has different variations in terms of position, posture (orientation), size etc. Object detection involves matching the model object with the instance of the object (target) in the image based on the properties of the object viz. color, shape, intensity, texture etc. Object is described using the specific property and this represents the feature vector of the object. Further the matching is performed using the feature descriptor. Automated object detection is being applied in varied fields like industry, medical imaging, defense applications, security and surveillance systems, office automations, mobile applications etc. Lot of research has been carried out in the field of object detection and variety of algorithms exists based on different properties of the object. The proposed system is designed for the rotation and scale invariant object detection based on the shape of the object. The technique is simple and computationally efficient. Unlike, most of the existing techniques which use classifier for detection and require lot of training data to train the classifier, the proposed system is based on detecting the object using single template (model) object.