摘要
针对红外图像行人检测中虚警率高的问题,提出了一种串联组合分类的红外行人检测方法。该方法利用感兴趣区域的宽高比和方向梯度直方图2种特征,采用串联组合方式实现分类识别。实验结果中检测率达到了90.5%,虚警率减少了43.7%,表明提出的方法可有效检测行人同时降低虚警。
An effective method using series combination classification was proposed to solve the problem of high false alarm ratio in infrared pedestrian detection. After obtaining the region of interest, two kinds of characters were computed, ratio of width to height and histograms of o- riented gradients （HOG）. The series combination strategy was used to detect the pedestrian. In the experimental results, the detection ratio reached 90. 5% and the false alarm ratio re- duced by 43.7 %. The results show that the proposed method is effective to detect the infrared pedestrian and reduce the false alarm ratio.
出处
《应用光学》
CAS
CSCD
北大核心
2012年第4期770-773,共4页
Journal of Applied Optics
