Object Recognition by Multi-Scale Color Local Binary Pattern

Abstract

This paper presents a multi-scale color local binary pattern based method for object recognition. Several advanced methods have been proposed for object recognition using deep neural networks. Although these methods offer high accuracy with many number of object classes, the high complexity of these methods requires large number of training examples and computational resources. This paper attempts to solve the object recognition problem for limited number of well known object classes to be used in practical scenarios. Color features are susceptible to change in varying illumination conditions. Shape features vary with the view angle. However texture features are more dependable. Multi-scale color local binary patterns are computed and artificial neural networks are trained. Experimental results on CALTECH 101, COIL 100 datasets indicate that the proposed method has good performance with low computational complexity.

Keywords: Object recognition, neural networks, classification, texture, local binary pattern

Cite this Article

Sunder R, Nagalinga Rajan A. Object Recognition by Multi-Scale Color Local Binary Pattern. Research & Reviews: Journal of Statistics. 2018; 7(1): 57s–61sp.