杂波环境下基于视听信息融合的目标跟踪  

Target tracking in clutter environment based on audio-video information fusion

作　　者：曹洁[1] 魏建勇[1] 

机构地区：[1]兰州理工大学电气工程与信息工程学院,甘肃兰州730050

出　　处：《兰州理工大学学报》2009年第3期85-89,共5页Journal of Lanzhou University of Technology

基　　金：甘肃省自然科学基金(0710RJZA060)

摘　　要：以监控系统为研究背景,充分利用场景中视听媒体间存在的天然时空相关性,将视觉信息和听觉信息有效地融合从而实现对目标的快速跟踪.利用机器视觉相关理论提取视觉运动特征,利用计算听觉场景分析技术抽取音频场景特征,建立视听信息特征级融合模型并进行联合场景事件判断.仿真结果表明,应用视听信息融合对目标进行跟踪,预测误差小于单独基于视觉的目标跟踪,同时听觉信息的引入有助于克服图像噪声.Taking surveillance system as investigation background, a fast target tracking was realized by means of full utilization of natural spatial-temporal correlation of audio and video media with each other in scene and effective information fusion of these two media. Visual features of motion were extracted by using the machine vision-related theory, and the technique of computational auditory scene analysis was used to extract the features of audio scene, so that a model of audio-video information fusion at feature level was set up and the joint scene events were judged. The simulation result show that when this audio-video information fusion was used to track the target, the forecast error was smaller than that when single video target tracking was used alone. Meantime, the introduction of auditory information was help full to overcome the image noise.

关 键 词：信息融合 特征提取 目标跟踪 卡尔曼滤波 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]