基于稳像技术的飞艇监控视频目标追踪

一般的目标追踪算法提取目标的颜色或轮廓特征,在图像区域内使用匹配算法完成对目标的追踪。由于飞艇容易受到气流影响,艇载相机平移误差会造成目标在视频的相邻帧间运动距离过大,传统目标追踪算法容易陷入到局部最优解而造成目标跟错或者丢失。该文提出了一种基于视频稳像的追踪方法,使用基于运动估计和混合滤波算法,首先处理视频使之平滑稳定,在此基础上利用人机交互选择目标并应用基于MeanShift的算法实现追踪。比较本文提出的算法和一般算法在飞艇视频目标追踪中的效果,结果表明:该方法在目标追踪中具有更高的准确率,同时满足实时性要求。实验证明了本文提出算法可以准确有效地处理飞艇视频目标跟踪问题。

Target tracking using airship supervision video based on video stabilization

Traditional target tracking algorithms make use of color or shape features to track the target using matching algorithms in areas near the target.However,airflow can cause large shift errors in the camcorder analysis due to large movements of the target in consecutive frames,resulting in tracking failures due to changes in the local maximum.This paper presents an adaptive Mean Shift tracking method based on video stabilization to track a target by applying motion estimation and mixed filtering to the video in advance.This system gives better tracking accuracy than other algorithms based on the Mean Shift method for real-time processing.Tests show effectiveness of this technique in the field for target tracking using airship supervision video.