面向监视视频实时分析的快速行人检测方法
为解决监视视频实时分析应用中行人检测效率低的问题,提出一种快速行人检测方法。首先,采用运动侦测方法提取运动区域,并结合行人检测要求对运动区域进行尺寸扩展、归一化和拼接操作;然后,在拼接图像上结合积分图快速提取各运动区域的Haar特征,并采用双支持向量机实现快速的特征分类;最后,结合包围盒相交策略进行帧间滤波,降低行人误检现象。实验表明,该方法不仅可以实时检测行人目标,而且检测错误率低于现有主流方法。...
Saved in:
| Published in | 计算机应用研究 Vol. 34; no. 4; pp. 1257 - 1260 |
|---|---|
| Main Author | |
| Format | Journal Article |
| Language | Chinese |
| Published |
合肥工业大学计算机与信息学院,合肥230009
2017
安庆职业技术学院电子信息系,安徽安庆246008%合肥工业大学计算机与信息学院,合肥,230009%安庆师范大学计算机与信息学院,安徽安庆,246133 |
| Subjects | |
| Online Access | Get full text |
| ISSN | 1001-3695 |
| DOI | 10.3969/j.issn.1001-3695.2017.04.068 |
Cover
| Summary: | 为解决监视视频实时分析应用中行人检测效率低的问题,提出一种快速行人检测方法。首先,采用运动侦测方法提取运动区域,并结合行人检测要求对运动区域进行尺寸扩展、归一化和拼接操作;然后,在拼接图像上结合积分图快速提取各运动区域的Haar特征,并采用双支持向量机实现快速的特征分类;最后,结合包围盒相交策略进行帧间滤波,降低行人误检现象。实验表明,该方法不仅可以实时检测行人目标,而且检测错误率低于现有主流方法。 |
|---|---|
| Bibliography: | For solving the problem of low efficiency of pedestrian detection for real-time analysis of surveillance video, this paper propsoed a fast pedestrian detection method. First, it used motion detection method to extract moving regions, and executed size extension, normalization and image mosaics according to the requirement of pedestrian detection. Then,it extracted Haar features of every moving regions fast by using integral image on mosaic image, and classified the features by using fast twin support vector machines. Finally,it executed inter-frame filtering by combining the policy of intersect bounding boxes. Experiments show that this method can not only detects pedestrians real-time ,but also has lower detection error rate than the current mainstream approaches. 51-1196/TP pedestrian detection; motion detection; support vector machines ; Haar; inter-frame filter; integral image; surveillance video Shu Guohui1,2, Ouyang Yiming1 , Su Benyue3 （ 1. School of Computer ＆ Information, Hefei University of Technology, |
| ISSN: | 1001-3695 |
| DOI: | 10.3969/j.issn.1001-3695.2017.04.068 |