基于改进单模高斯模型的运动目标检测算法  

Moving target detection based on improved single-mode Gaussian model

作　　者：门瑜 郑娟毅[1] 李萌[1] 

机构地区：[1]西安邮电大学通信与信息工程学院,陕西西安710121

出　　处：《电视技术》2016年第4期18-21,64,共5页Video Engineering

基　　金：国家自然科学基金项目(11401469);陕西省工业攻关项目(2013K06-07)

摘　　要：在视频交通车辆目标检测中,阴影问题是影响其检测准确性的关键问题之一。为了解决这个问题,提出了一种结合单模高斯模型和背景差法的运动目标阴影检测方法。首先针对传统单模高斯模型提出了一种自适应学习率和选择性差值更新背景相结合的方法,加快了背景模型的初始化速度,同时结合背景差法对阴影部分进行检测与去除。实验结果表明,该方法能够较好地去除车辆的阴影,提高了检测的准确性。In the video transport vehicle target detection, shadow problem is one of the key issues that affects the detection accuracy. To solve this problem, a shadow moving object detection method combined with single-mode Gaussian model and background subtraction is proposed. Firstly, according to the traditional single-mode Gaussian model, the method of an adaptive learning rate and selective difference update background is proposed, and the speed of background initialization model is accelerated, mean- while, combined with background subtraction shadow detection and removal section. Experimental results show that this method can remove the shadow of the vehicle and improve the accuracy of detection.

关 键 词：运动目标检测 单模高斯模型 背景差法 阴影检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]