【摘要】 针对红外图像的特点和行人形状特征提出一种快速高效的行人检测算法。采用基于直方图统计的显著性映射算法获取红外图像的显著图(SM),统计出SM中关注点(AP)的分布,确定了自适应分割阈值;针对行人姿势的多样性,结合先验概率构建基于形状的级联模板树,在分割图像上根据匹配值确定行人的位置。选取3个公开数据集对比几种行人检测算法。实验结果表明,所提算法在精度和检测速度方面都有明显优势。更多还原

【Abstract】 An effective approach for pedestrian dectecting is presented in this paper considering the characteristics of infrared images and human shapes.A method based on the statistics histogram of contrast is used to calculate saliency maps(SM)of infrared images.Attention points(AP)extracted from SM are used to decide the segmentation threshold.The hierarchical part-template match tree is built combined with prior probabilities and varieties of human posture.After matching template-tree with segmented results,pedestrians are presented.Experiment results of the approach are compared with that of other methods on three open datasets.The results show that the proposed algorithm has the high precision with less time.更多还原