面向人群分析的智能监控视频检测研究

Research on Intelligent Surveillance Video Detection for Crowd Analysis

文章提出一种新型的多人检测模型,用于室内外监控环境中的人物计数和异常事件检测。首先,该模型使用逆变换和中值滤波器提取人体轮廓,降低了计算和处理各种复杂监测情况的成本。其次,使用具有Jaccard相似性和归一化互相关的卡尔曼滤波器技术,实现连续帧中的人物跟踪。最后,通过高斯映射进行聚类,用于异常事件检测。实验结果表明,该模型在计数准确率和检测率方面均具有较好的性能。

This paper proposes a new multi person detection model for character counting and abnormal event detection in indoor and outdoor monitoring environments.Firstly,the model uses inverse transform and median filter to extract human body contours,reducing the cost of calculating and processing various complex monitoring situations.Secondly,a Kalman filter technique with Jaccard similarity and normalized cross correlation is used to achieve character tracking in consecutive frames.Finally,clustering is performed through Gaussian mapping for abnormal event detection.Experimental results show that,the proposed model exhibits better performance in terms of counting accuracy and detection rate.