NobleBlocks

Research indexJournal article

3D Object Recognition Framework

Patil, Lakamanhalli

1 May 2018

NobleIDark:/48914/w1/20260525/E83D4D7ADOI10.1109/rteict42901.2018.9012580

Paywalled

The full text is at the publisher

You can read the abstract here, or open it at the publisher.

Read at the publisher

Abstract

Object recognition is one of the inherent characteristics which humans own since adolescence. In reality, it is astonishing to see that humans can recognize any object at a single glance. Machines however, find it difficult to recognise objects since the object appearance may have variation due to photometric effects, clutter, occlusions or rotation. But, all these variations are instantaneously handled by human brains. In spite of the fact that the objects may vary depending on point of view or any changes in size or scale, humans can smoothly identify the object. This task is yet a major problem for machines. Hence, there is a need to develop a model which can efficiently and effectively recognize 3D objects. This paper proposes a 3D object recognition framework in order to solve the problem faced by computer vision systems. In the proposed model, key points are extracted using Intrinsic Shape Signature (ISS). The geometric information surrounding these points is determined by using Fast Point Feature Histogram (FPFH) descriptor. The resulting feature matrix is the local feature of an object. Feature sets are formed by using Efficient Match Kernel (EMK) and these feature sets are used as weights in Support Vector Machine (SVM) classifier to classify the object in a scene to the category in a training set.

Discussion

Citations

Topics

More

Suggest a correction or report