期刊名称：International Journal of Computer Science and Information Technologies
电子版ISSN：0975-9646
出版年度：2016
卷号：7
期号：1
页码：407-412
出版社：TechScience Publications
摘要：Shape is one of the high level features that play animportant role in the object recognition and perception.Object shape features provide a powerful clue to objectidentity. In this paper, we have implemented the approach bycombining three feature descriptor i.e. edge histogramdescriptor, geometric moment, and Sobel edge detectortechniques recognize the objects in the images that is invariantwith the changes , scaling, rotation, and orientation. Sinceedges play an important role in image perception and it isfrequently used as a feature descriptor in image retrieval sowe select the Edge Histogram Descriptor, EHD, as a featurevector which represents the spatial distribution of five types ofedges. We also select geometric moment invariant as anothershape feature vector which is extensively used to extort globalfeatures, can improve the previously recognition rate to asignificant measure. Due to Sobel operator’s smoothing effect,it is less sensitive to noise present in Images. On the otherhand, smoothing affects the accuracy of edge detection. Sonext we select Sobel edge detection as a third feature vectorthat extract the shape features. Finally, we combined the allthree shape features that form the 3 dimensions featurevectors of the entire image. Experiment results conclude thatthe proposed scheme has a very good performance in respectof the precision recall.
