The Multi-Orientation Target Recognition Method Based on Visual Attention

Synthetically utilizing image visual attention and Support Vector Machine (SVM) classification method, a multi-orientation target recognition algorithm was proposed to detect multi-orientation targets in images. Firstly, according to human visual system, the saliency image was get rapidly using visual attention to improve the efficiency. Secondly, the Histogram of Oriented Gradients (HOG) features described the shape features of target. Then, the angular field of view to targets was divided into several parts for solving the samples variety according to the pose angle. In every divided field SVM classifier was used to recognize the multi-orientation targets. Experimental results show that the multi-view target recognition method proposed by this paper is effective and reliable.