基于智能Agent的车辆自适应跟踪技术  

Adaptive Technology for Vehicles Tracking Based on Intelligent Agent

作　　者：曲巨宝[1] 

机构地区：[1]武夷学院数学与计算机系,福建武夷山354300

出　　处：《西南交通大学学报》2011年第4期626-632,共7页Journal of Southwest Jiaotong University

基　　金：福建省自然科学基金资助项目(2007J0189);福建省教育厅科技资助项目(JA09240);武夷学院智能计算网格科研团队计划项目资助(2009)

摘　　要：为了研究复杂环境下快速移动车辆目标检测与跟踪问题,提出了基于知识库的智能Agent自适应图像分割与滤波算法,建立了帧间差异积累动态矩阵自适应背景模型,在跟踪过程中,设计了改进的SSD算法预测初始迭代点,根据Jensen不等式推导了具有自适应核窗宽迭代更新的M eanSh ift算法,实现了对视频车辆目标的自适应智能跟踪.实验结果表明,该算法能有效、准确地跟踪视频中的运动目标,自适应能力强;与其他算法比较,跟踪误差降低了54.4%,平均跟踪时间延长了41.3%.To study the problem of detecting and tracking fast-moving vehicles in complex circumstances,an adaptive algorithm for image segmentation and filtering was proposed using intelligent Agent based on knowledge database,and an adaptive background model was built through the dynamic matrix of accumulated frame differences.In the tracking process,an improved SSD（sum of squared differences） algorithm was designed to forecast the initial iteration points.According to the Jensen inequality,a MeanShift algorithm for iterative updating of the adaptive kernel-bandwidth was derived to achieve the adaptive intelligent tracking of moving vehicles in videos.The Experimental results shows that the algorithm can track the moving targets in videos effectively and accurately,and has a strong adaptive ability.Compared with other existing algorithms,the tracking error is reduced by 54.4%,and the average tracking time is extended by 41.3%.

关 键 词：智能AGENT 跟踪 自适应 MEANSHIFT算法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]