基于反馈背景模型的城市道路交叉口前景目标检测

为准确检测城市道路交叉口监控视频中缓慢行驶或短时停留的前景目标,提出了一种基于前景目标反馈的背景模型检测方法.首先基于观测样本的像素值构建背景模型,利用计数器观测像素点检测为前景或背景的次数并描述当前场景的交通状态和稳定性,其次根据场景自适应阈值判断当前像素点为前景点或背景点,最终通过交通状态和场景的稳定性更新背景模型.采用基于真实的交叉口视频场景对算法的有效性进行了定性与定量分析.实验结果表明,该算法在复杂的城市道路交叉口场景中检测出缓慢行驶或短时停留车辆的性能优于其他方法,同时能够满足城市道路交叉口智能视频监控实时性和准确性的要求,为交叉口前景目标的行为分析奠定了基础.

Foreground Objects in Surveillance Video of Urban Traffic Intersection Using Feedback Background Subtraction Model

To accurately detect the slow motions or temporarily stopped foreground objects in surveillance video of urban traffic intersection，a background subtraction model is proposed using the feedback of foreground objects. The background template is built based on the observed pixel values and each pixel is assigned counters to describe the current traffic state and the stability of a pixel. The foreground decision depends on an adaptive threshold, and background model update is based on the feedback current traffic state and the stability. The overall results obtained with the real- world urban traffic videos are presented to demonstrate that the proposed method achieves better performance of both qualitative and quantitative evaluation than other state- of- the- art methods in the slow motions or temporarily stopped objects traffic scenario. This method satisfies the requirement of the real-time and accuracy of the intelligent video in the urban traffic intersection, and the foundation for intelligent video analysis is laid.