State-of-the-Art Object Recognition Techniques: A Comparative Study

Abstract

Object recognition has a remarkable contribution in the field of computer vision. It has many areas of application like security, human–computer interface, industrial inspection and automation, etc. This paper presents the distinct object recognition approaches like feature-based method, appearance-based method and artificial neural network. Further, the various state-of-the-art algorithms like Scale Invariant Feature Transform (SIFT), Speeded-Up Robust Features (SURF), Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Convolutional Neural Network (CNN) of these approaches are introduced along their pros and cons. Finally, we conclude with a comparison of these algorithms on the basis of robustness (in terms of rotation, illumination, occlusion and speed), complexity (computation load and memory usage) and accuracy.