期刊名称：International Journal of Image Processing (IJIP)
电子版ISSN：1985-2304
出版年度：2014
卷号：8
期号：5
页码：278-293
出版社：Computer Science Journals
摘要：This paper discusses an approach for object detection and classification. Object detection approaches find the object or objects of the real world present either in a digital image or a video, where the object can belong to any class of objects. Humans can detect the objects present in an image or video quite easily but it is not so easy to do the same by machine, for this, it is necessary to make the machine more intelligent. Presented approach is an attempt to detect the object and classify the same detected object to the matching class by using the concept of Steiner tree. A Steiner tree is a tree in a distance graph which spans a given subset of vertices (Steiner Points) with the minimal total distance on its edges. For a given graph G, a Steiner tree is a connected and acyclic sub graph of G. This problem is called as Steiner tree problem where the aim is to find a Steiner minimum tree in the given graph G. Basically the process of object detection can be divided as object recognition and object classification. A Multi Scale Boosted Detector is used in the presented approach, which is the combination of multiple single scale detectors; in order to detect the object present in the image. Presented approach makes use of the concept of Steiner tree in order to classify the objects that are present in an image. To know the class of the detected object, the Steiner tree based classifier is used. In order to reach to the class of the object, four parameters namely, Area, Eccentricity, Euler Number and Orientation of the object present in the image are evaluated and these parameters form a graph keeping each parameter on independent level of graph. This graph is explored to find the minimum Steiner tree by calculating the nearest neighbor distance. Experimentations are carried out using the standard LabelMe dataset. Obtained results are evaluated based on the performance evaluation parameters such as precision and recall.
