摘要
无人机侦察任务中的目标定位,需要高效地获取机载视频数据,并采用实时性鲁棒性较强的目标检测算法;针对上述要求,选择在DirectShow框架下实现具有图像处理功能的自定义转换过滤器作为处理平台;同时提出了采用基于卢卡斯卡那得光流和连续自适应均值平移互补的方法进行目标检测定位;经过实验测试,结果证明了该方案在满足实时性要求的同时还能够解决图像色度信息较差、目标图像较小以及目标区域比较平滑时的检测定位问题,是一种针对战场目标定位的有效方法。
Target location for UAV （Unmanned Aerial Vehicle） scout task involves an efficient access of video data and a real-time robust target detection algorithm. Accroding to these requirements, a process platform was chosen which was built by realizing a customed transform-filter with image process function in DirectShow architecture; and the Lucas-Kanade optical flow and continuously apative Mean-Shift algorithm were used together for target location in this project. The experiment results demonstrate that several detection problems can be resovled , such as poor hue, small area and smooth texture in target area; and the real-time performance can be guaranteed at the same time with this efficient location method for battlefield targets.
出处
《计算机测量与控制》
CSCD
北大核心
2010年第4期900-901,905,共3页
Computer Measurement &Control
