基于局部轮廓特征的无人机头部检测跟踪算法  

Detection and tracking algorithm for head of UAV Based on local contour features

作　　者：王坤[1] 王磊[1] 游安清[1] 

机构地区：[1]中国工程物理研究院应用电子学研究所,四川绵阳621900

出　　处：《光学技术》2011年第2期178-182,共5页Optical Technique

摘　　要：针对无人机的结构特点,提出了一种基于目标轮廓提取、轮廓自动分段、头部定位以及头部模板匹配四个模块的目标头部检测跟踪算法。首先对第一帧图像提取目标轮廓,并通过计算轮廓点的变化率,对轮廓进行自动分段,得到目标的4个局部轮廓点列,进而根据头部轮廓点列的特点,完成头部轮廓筛选,然后利用基于目标中轴的二次曲线拟合,实现头部顶点的检测,并在轮廓图中提取头部模板。对后续帧,通过与前一帧进行头部模板匹配,完成目标图像序列的头部检测跟踪。实验表明,该算法能够有效地实现无人机头部的检测跟踪。An algorithm for detection and tracking for head of UAV based on contour extraction,automatic contour section,head point detection and head template matching with its structure feature of UAV image sequence is proposed.Firstly,the contour of UAV of first frame is extracted,and using by computing the contour point＇s rate of change,the contour is segmented automatically.The segments can be classified into 4 kinds,and then the head segment is selected according to its feature.These points on head segment can be fitted into twice curves using target＇s central axis,and then the head top-point and the head template are extracted.According to matching the head template with the contour image of follow-up frames,it can implement the head detection and tracking with image sequence.The experiment results indicate that this algorithm is effective on detection and tracking for UAV＇s head.

关 键 词：头部检测 轮廓分段 局部轮廓筛选 曲线拟合 特征匹配 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]