智能监控中基于头肩特征的人体检测方法研究

针对传统监控系统存在的不足,研究了智能监控中人体目标的自动检测,提出了运动目标头肩模型提取和支持向量机头肩模型验证相结合的新方法,该方法将差分图像的边缘检测与轮廓跟踪相结合,有效地消除了目标影子干扰的影响;利用人体局部形状特征区分人体目标,解决了实际应用场合中人体易受到遮挡的问题;基于支持向量机(SVM)的分类器克服了传统方法在小样本条件下容易欠学习与过学习的问题.实验结果表明,该方法具有较强的鲁棒性和较高的正确率,在智能监控系统中能自动检测运动人体目标,为智能监控中人体目标自动检测和跟踪提供理论和技术基础.

Human detection based on head and shoulder feature in intelligent surveillance system

Aimed at the shortcomings of the traditional visual surveillance system, this paper proposed a series of methods to solve the difficulty in detecting human under complex background. The object was abstracted accurately by combination of edge detection of the difference image and contour tracking to bypass the object shade. Human part shape analysis solved the problem of the human image being easily occluded in practical application. Classifier based on Support Vector Machine(SVM) overcame the disadvantages of overfitting in traditional methods. Experimental results show that this method is strongly robust and highly accurate; and can detect human automatically under complex background; and provides theoretical and technological base for object detection and tracking in the intelligent surveillance system.