基于声图像序列的快速目标跟踪算法  

A Fast Object Tracking Algorithm Based on Sonar Image Sequences

作　　者：田晓东[1] 周德超[1] 刘忠[1] 

机构地区：[1]海军工程大学电子工程学院,湖北武汉430033

出　　处：《计算机仿真》2007年第7期5-8,共4页Computer Simulation

摘　　要：针对声图像具有对比度低、不同时刻目标回波变化大等特点,提出了一种基于卡尔曼滤波和数据关联算法的快速目标跟踪算法。该算法首先对声图像序列进行高斯平滑和自适应阈值分割处理,在此基础上建立卡尔曼滤波目标跟踪模型和数据关联目标匹配算法,分析计算了可跟踪的目标速度上限值。仿真结果表明,该算法可较准确地实现目标跟踪,允许的目标速度可满足实际应用的需要。同传统文献算法相比,该算法具有实时性好、鲁棒性强,能更好地适应目标分裂或合并等情形。Aiming at the characteristics of sonar image such as low - contrast and intensity of object varies greatly at different instant, a novel fast object tracking algorithm based on Kalman filter and data association is proposed. First, this algorithm utilizes Gaussian filter and adaptive thresholding method to preprocess sonar image. Secondly, a tracking model based on Kalman filter and data association method is presented. The upper - limit of object＇ s velocity is analyzed. Simulation result indicates that the algorithm presented in this paper can track object in sonar image precisely and meet the application requirement. Compared with traditional methods provided in literatures, this algorithm has characteristics such as real -time, great robustness.

关 键 词：目标跟踪 声图像序列 卡尔曼滤波 数据关联 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]