Object Recognition via Classifier Interaction with Multiple Features
Jie Tang, Gongjian Wen · 2016
In this paper, a robust object recognition method via classifier interaction with multiple features is proposed to recognize an object in dynamic conditions that include illumination changes, pose variations, and occlusions. To account for the better recognition effect, multiple classifiers with different features are utilized, and each of which will get satisfactory result by classifier selection and interacting with each other. To integrate multiple classifiers for accurate object recognition, we propose two solutions: (1) the classifier selection, and (2) classifier interaction modules within a Bayesian framework. The experimental results demonstrate that our proposed method is robust and highly active in the field of object recognition.