基于视觉传达的多帧视频图像邻域跟踪仿真

Multi-frame Video Image Neighborhood Tracking Simulation Based on Visual Communication

在处理复杂多帧视频图像时,传统的基于SURF特征的图像邻域跟踪方法暴露出目标邻域精度不高、边界不够平滑且无法满足实时处理要求的情况.针对上述情况提出了一种基于视觉传达的多帧视频图像邻域跟踪方法.方法首先通过选取适当的叠加尺度,将邻域特征描述的向量与适当的叠加尺度进行完全叠加,完成去除冗余特征点和离散点处理,再对叠加区域中出现的目标给予权值,最后将权值与显著性加权最小二乘图像匹配方法结合完成多帧视频图像的邻域跟踪.实验表明:提出的方法不仅可以提高多帧视频图像的特征提取与匹配的精准率,而且降低了在匹配过程中出现的噪声问题,证明了上述方法相对于现有的邻域跟踪方法有很大的优越性,为下一步的多帧视频图像领域跟踪提供了精确的方法.

When processing complex multi-frame video images, the traditional image neighborhood tracking method based on SURF feature leads to low accuracy of target neighborhood, and the boundary is not smooth enough, so that the image cannot be processed in real time. Therefore, a multi-frame video image neighborhood tracking method based on visual communication was presented. Firstly, this method selected appropriate superimposed scales, and then vectors described by neighborhood features were completely superimposed with appropriate superimposed scales, and thus to remove the redundant feature points and discrete points. Secondly, our method gave the weights to the targets appearing in the superimposed region. Finally, the method combined the weights with the saliency weighted least squares image matching tracking method to track neighborhood of multi-frame video image. Simulation results show that the proposed method not only improves the accuracy of feature extraction and matching precision of multi-frame video images, but also reduces the noise in matching process. Compared with the existing neighborhood tracking methods, the proposed method has great advantages, which provides an accurate method for next neighborhood tracking.