摘要
大型公共建筑内人群数目及分布的在线监测是有效控制和疏散客流、保障人员安全的重要依据之一.利用公共建筑内现有的闭路电视监视系统,通过计算机视觉技术实现人群数目的自动识别是目前国外普遍采用的一种方式.文中提出了一种基于RBF神经网络的复杂场景人群目标的识别算法,利用包含行人数目信息的前景图像的投影曲线等特征数据,通过训练好的RBF神经网络直接得到该前景图像中包含的人群数目.与其他算法相比,该算法具有较高的识别准确率,在一定误差范围内可以达到较好的效果.
Monitoring real-time pedestrians＇ information on-line in large public buildings is very important to control and evacuate these pedestrians. By existing closed-circuit television surveillance systems in public buildings, automatic identification of crowds based on computer vision technology is commonly used abroad. In this paper, pedestrians＇ data reorganization algorithm based on RBF neural network is proposed. This algorithm can recognize extractive foreground image and collect pedestrians＇ data including their number and density. Compared with other algorithms, the presented method is low timeconsuming and high precise.
出处
《北京交通大学学报》
CAS
CSCD
北大核心
2009年第4期29-33,共5页
JOURNAL OF BEIJING JIAOTONG UNIVERSITY
基金
北京市自然科学基金资助项目(9052007)
