室内单目监控视频动态前景提取方法研究  

Research on Dynamic Foreground Extraction Method for Indoor Monocular Surveillance Video

作　　者：刘冰清 

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海

出　　处：《建模与仿真》2023年第3期2390-2399,共10页Modeling and Simulation

摘　　要：针对室内单目监控视频光照不均等原因产生的阴影干扰和设备、环境因素产生的抖动问题,本文提出了一种特征定位与改进帧差法融合的动态前景目标提取方法。首先,基于像素级消减动态阴影特征并干扰使用高斯滤波进行降噪处理以减少监控视频冗余特征信息。其次,使用特征匹配算法获取连续两帧图像的差分,初步捕获动态前景目标。然后,使用自适应阈值和形态学处理方法改进帧差法精确提取前景动态目标。最后,实验验证方法的有效性和精准性,本文室内单目监控视频动态前景目标提取算法准确度达到92.2%,有效消除室内单目监控视频中的阴影和抖动干扰现象。Aiming at the shadow interference caused by uneven illumination of indoor monocular surveillance video and the jitter problem caused by equipment and environmental factors, this paper proposes a dynamic foreground target extraction method based on feature location and improved frame dif-ference method. Firstly, dynamic shadow features are reduced at the pixel level and Gaussian fil-tering is used for noise reduction to reduce redundant feature information of surveillance video. Secondly, the feature matching algorithm is used to obtain the difference between two consecutive frames of images to initially capture the foreground dynamic target. Then, the adaptive threshold and morphological processing method are used to improve the frame difference method to accu-rately extract the foreground dynamic target. Finally, the effectiveness and accuracy of the method are verified by experiments. The accuracy of the dynamic foreground object extraction algorithm in indoor monocular surveillance video reaches 92.2%, which effectively eliminates the shadow and jitter interference in indoor monocular surveillance video.

关 键 词：监控视频 自适应阈值 前景目标 形态学处理 特征定位 提取算法 高斯滤波 光照不均 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]