燕山大学学报2016，Vol.40Issue(1)：66-73,8.DOI:10.3969/j.issn.1007-791X.2016.01.011
基于稀疏表示分类行人检测的二级检测算法
Secondary detection algorithm of pedestrian detection based on sparse representation classification
摘要
Abstract
Using method of fast feature pyramids to detect pedestrains in static images there are a lot of false positive detected win⁃dows in the result which reduce the accuracy of pedestrian detection. To decrease the amount of false positive detected windows a method of two stage cascade of pedestrian detection is proposed based on the result of using fast feature pyramids algorithm to detect pedestrian secondary detection which bases on sparse representation classification is added to further decrease the number of false positive detected windows.Overcomplete dictionaries with improved HOG and improved HOG+LBP feature which are extracted from positive and negative samples are set up at the beginning of detection.During the detection stage sparse coefficient is solved by local weighted sparse representation.Whether detection windows contain pedestrians can be judged by a ratio which sums the positive and negative coefficient separately.After added a secondary detector the number of false positive detected windows decreases dramatical⁃ly log⁃average miss rate goes down and average precision is improved.Also processing time meets the requirements of real⁃time de⁃tection.关键词
行人检测/稀疏表示/局部区域权重/HOG/LBPKey words
pedestrian detection/sparse representation/local area weight/HOG/LBP分类
信息技术与安全科学引用本文复制引用
胡春海,张凯翔,范长德..基于稀疏表示分类行人检测的二级检测算法[J].燕山大学学报,2016,40(1):66-73,8.基金项目
国家自然科学基金资助项目 （）
