基于稀疏表示的两级级联快速行人检测算法

Two Stage Cascade Fast Pedestrian Detection Algorithm Based on Sparse Representation

针对复杂背景下的行人检测问题,提出一种两级级联的快速行人检测算法。第一级采用竖直方向的边缘对称特征和基于行人先验知识的弱分类器,排除大部分非行人区域。第二级采用梯度方向直方图特征和基于LC-KSVD字典学习的稀疏表示分类算法,对剩余区域进行精确检测。实验结果表明,该算法在保证检测精度的同时缩短了行人检测的时间,并且对遮挡情况有较好的鲁棒性。在INRIA数据库上每幅图像平均检测时间仅为69ms,对数平均漏检率为38%,较CENTRIST+C4算法和HOG+SVM算法的漏检率有所降低,并提升了检测速度。

According to the problem of fast pedestrian detection in complex background,this paper proposes a two stage cascade fast pedestrian detection algorithm. At the first stage,most of non-pedestrian areas can be excluded by using the vertical edge symmetrical feature and the weak classifier based on pedestrians prior knowledge. During the second stage, the remaining areas are further accurately detected with Histograms of Oriented Gradient（ HOG） features and the sparse representation classification algorithm based on LC-KSVD dictionary learning. Experimental results show that the algorithm ensures the detection accuracy and shortens the pedestrian detection time. It also has good robustness to occlusion. In the INRIA database,the average time required for each image is only 69 ms,the logarithmic mean missing rate is 38%, compared with CENTRIST ＋ C4 algorithm and HOG ＋ SVM algorithm, the relative missing rates are decreased,and the relative detection speeds are increased.