一种基于非对称决策融合的行人检测方法

A Pedestrian Detection Method Based on Asymmetric Decision Fusion

行人检测一直是计算机视觉中的热门话题,其重要应用包括视觉监视、机器人技术和汽车驾驶等。而当行人暂时保持静止时,运动信息将消失,运动特征与外观特征均不稳定,故一般特征融合方法在该情况中无法获得良好效果。针对于此类问题,论文提出了一种非对称特征融合方法来处理此类不稳定信息源的融合问题,经实验表明,该方法具有较好的检测效果。

Pedestrian detection has been a hot topic of computer vision due to its important applications such as visual surveillance,robotics and automotive driving.As the motion information will disappear when the pedestrian temporarily keep stationary,the motion feature is unstable compared with the appearance feature makes the general fusion method not appropriate in the decision fusion of appearance and motion feature.Therefore,this paper proposes an asymmetric fusion method to handle the fusion problem with one unstable information source and it is proved that this method works well in the real pedestrian detection.