An efficient object recognition based on Gabor transform and LBP variance

摘要

Recognizing objects from arbitrary aspects is always a highly challenging problem in applied engineering and computer vision fields. At present, most existing algorithms mainly focus on specific viewpoint detection. Hence, in this paper we propose a novel recognizing model, which combines Gabor transform with LBP variance to handle the problem of different viewpoints and pose changing. Then, the images of inaccurate recognizing are evaluated by learning and fed back the detector to avoid the same mistakes in the future. The principal idea is to extract intrinsic viewpoint invariant features from the unseen poses of object, and then to take advantage of these features to support recognition. Compared with other recognition models, the proposed approach can efficiently tackle the multi-view problem and promote the recognition performance. For a quantitative evaluation, this novel algorithm has been tested on two benchmark datasets such as Caltech 101 and PASCAL VOC 2011datasets. The experimental results validate that our approach can recognize objects more precisely and outperforms others single view recognition methods.