基于MFFK和Mean-Shift的红外船舶目标跟踪算法研究  

MFFK and Mean-Shift combined algorithm for infrared ship target tracking

作　　者：徐晓波[1] 李卫斌[1,2] 

机构地区：[1]西安科技大学电气与控制工程学院,西安710054 [2]咸阳师范学院信息工程学院,陕西咸阳712000

出　　处：《计算机工程与应用》2013年第5期144-146,204,共4页Computer Engineering and Applications

基　　金：国家自然科学基金重点项目(No.60736007);陕西省自然科学基金项目(No.2007D22);陕西省教育厅自然科学基金项目(No.08JK479);全国博士后基金项目(No.20070411140)

摘　　要：当前内河船舶配备的导航设备,如雷达、无线电信号接收避险等,无法在各种恶劣气候条件下准确获取目标的信息,不能满足可靠的船-船、船-桥避碰要求。采用基于MFFK分形特征与Mean-Shift方法相结合的红外船舶目标检测跟踪技术,在其运动过程中实时采集来自船运动前方以及两侧环境情况的视频文件,同时将其转化为单帧的实时红外图像,通过MFFK参数将环境与可疑人造目标分离开,然后采用基于Mean-Shift的目标跟踪技术进行目标定位,跟踪,通过对连续红外图像高速比对,在其运动轨迹发生变化可能产生事故之前发出警报。实验结果表明,该算法能够应付场景的各种变化以及多运动船舶目标交错遮挡等情形,算法具有鲁棒性。Since the navigation equipment for inland river ships, such as radar, radio reception hedge etc, as target information can not be accurately obtained in various heavy climate condition, it can not satisfy the requirement of reliable ship collision avoidance. This paper employs infrared ship target detection tracking technique based on MFFK fractal characteristics and Mean-Shift combined algorithm. The real-time motion infrared video is acquired along the ship direction and on both sides. The single frame picture is extracted to separate suspicious man-made objects from the background environment via MFFK parameter. Mean Shift method is employed for target tracking. The experimental results show that all kinds of changes and situation like scene movement target crisscross can be dealt by this algorithm. The algorithm is robust.

关 键 词：船舶碰撞 检测跟踪 多尺度分形特征(MFFK) 均值漂移(Mean-Shift)算法 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]