基于局部多层次特征的嵌入式视频全景拼接

Embedded Video Panorama Stitching Based on Local Multi-level Feature

嵌入式视频全景拼接技术是透视装甲的关键技术之一,在恶劣的战场环境中,传统的嵌入式视频全景拼接算法存在实时性不高,错位、重影明显等问题。针对以上问题,提出了一种基于局部多层次特征匹配的嵌入式视频全景拼接算法。选择重叠区域对图像进行特征提取来有效减少无用特征点提取的数量,提高实时性;然后计算图像不同层次的单应性矩阵H,避免单一H矩阵造成的错位问题;最后基于能量关系来查找最佳拼接线,消除运动场景下的重影缺陷。实验结果表明,该方法能够满足战场运动场景下视频全景拼接的实时性要求,拼接后的全景图像没有明显的错位、重影缺陷。

Embedded video panorama stitching technology is one of the key technologies for perspective armor.However,in the harsh battlefield environment,the traditional embedded video panorama stitching algorithm has problems of low real-time performance,misalignment,and ghosting.In view of the above problems,an embedded video panoramic decomposition algorithm based on local multi-level feature matching is proposed.First,select the overlapping area to perform feature extraction on the image to effectively reduce the number of useless feature points and improve real-time performance;then calculate the homography matrix H at different levels of the image to avoid the misalignment problem caused by the single H matrix;and finally based on the energy relationship Find the best dividing line to eliminate ghosting defects in motion scenes.Experimental results show that this method can meet the real-time requirements of video panorama stitching in battlefield sports scenes.The stitched panorama image has no obvious misalignment and ghosting defects.