【摘要】 为了克服传统行人检测方法计算复杂度高、耗时长的问题,提出一种基于多重分形谱的行人检测方法.采用多重分形理论分析行人图像并计算对应多重分形谱,提取多重分形谱特征,选取支持向量机(support vector machine,SVM)二值分类器实现行人滑窗检测.将本文算法应用于INRIA数据库中,并与基于HOG、HOGNMF特征的算法进行比较.实验结果表明:本文算法检测率为86.59%,虚警率为7.75%,有较高的检测率和较低的虚警率.更多还原

【Abstract】 To overcome the high computational complexity and time-consuming of traditional pedestrian detection, a pedestrain detection method using multifractal spectrum(MFS) was proposed. The multifractal theory was used for analyzing the pedestrian image and calculating the corresponding MFS. The MFS feature was extracted, and support vector machine(SVM) was selected to achieve slide-window detection for pedestrian. The proposed method was compared with the algorithms based on HOG and HOG-NMF on INRIA database. The experimental results showed that the detection accuracy of the proposed algorithm was 86.59%, the false alarm rate was7.75%, the algorithm has higher detection rate and lower false alarm rate.更多还原