改进的ViBe算法及其在交通视频处理中的应用

发布时间：2018-01-14 09:17

本文关键词：改进的ViBe算法及其在交通视频处理中的应用　出处：《光学精密工程》2017年03期 　论文类型：期刊论文

【摘要】：基于ViBe算法的运动目标检测具有很高的处理效率,但在复杂背景下也存在缺陷,比如不能有效抑制运动目标的残影和鬼影区域、易受噪音干扰,在交通视频的处理中更为明显。鉴于ViBe算法的这些不足,本文对ViBe原算法提出了改进,结合像素生命长度对鬼影或残影像素进行判断,利用二次更新策略提高消除速度。根据像素值变化引入最大类间方差法,将原有的固定像素阈值变为动态阈值,以提高交通视频中车辆目标检测的抗干扰能力。实验结果表明,在3组不同交通场景视频中,改进算法能够在较少帧数内消除目标的残影或者鬼影。对结果的量化分析也证明了改进算法由于应用了动态阈值,使目标检测的精准度和整体性都有所提高。
[Abstract]:Moving target detection based on ViBe algorithm has high processing efficiency, but it also has some defects in complex background, for example, it can not effectively suppress the residual and ghost regions of moving targets, so it is vulnerable to noise interference. It is more obvious in traffic video processing. In view of these shortcomings of the ViBe algorithm, this paper proposes an improvement to the original ViBe algorithm, combining with the pixel life length to judge the ghost or residual pixels. The quadratic update strategy is used to improve the elimination speed. According to the change of pixel value, the maximum inter-class variance method is introduced, and the original fixed pixel threshold is changed into dynamic threshold. In order to improve the anti-jamming ability of vehicle target detection in traffic video, the experimental results show that in three groups of different traffic scene video. The improved algorithm can eliminate the residual or ghost images of the target in less frames. The quantitative analysis of the results also proves that the improved algorithm can improve the accuracy and integrity of target detection because of the application of dynamic threshold.
【作者单位】：南昌大学信息工程学院;天津工业大学计算机科学与软件学院;
【基金】：国家自然科学基金资助项目(No.61302127) 江西省自然科学基金重大资助项目(No.20161ACB20004)
【分类号】：TP391.41;TN948.6
【正文快照】： 1引言视频监控是现代社会安防系统的重要组成部分,视频监控的处理越来越趋向于智能化和自动化[1],人们需要其能够自动采集信息、智能分析和处理数据以及自动反馈结果[2]。而交通监控较室内监控环境更为复杂,背景多是多模态场景,噪音较多,视频中运动车辆的速度、大小均不相同。

【相似文献】