【摘要】 提出了融合自底向上和自顶向下模式的视觉注意机制的改进算法。该算法使用以完整轮廓信息作为物体识别标志的方式,几乎不需要先验知识,改进了自顶向下算法的需要大量先验知识的弱点;同时由于将物体轮廓作为的一个识别和预检测的重要参考信息,改进原视觉机制中的仅使用三个显著图,没有整体意识,从而致使识别范围过大的弱点。实验证明该方法提升了搜索预检测目标的准确度。更多还原

【Abstract】 An improved integration model combining the bottom-up and top-down cues for visual attention is proposed.It uses an integrated outline as the symbol for object recognition and overcomes the shortcomings of the top-down visual attention method that needs excessive a priori knowledge.Furthermore,adopting the object outline as the main character for recognition and pre-detection also overcomes the drawbacks existing in the bottom-up visual attention model that only uses three saliency maps from the original picture,resulting in an overextended recognition region.Experimental results show that compared to the original algorithm,the proposed algorithm is more effective for target pre-detection.更多还原