基于局部单应性矩阵的图像拼接与定位算法研究

针对单个摄像机的视野范围有限导致在大场景下监控效果不够理想的问题,提出了一种改进的图像拼接与目标定位算法。该算法以多个摄像机获取的具有共视区域的监控图像为基础,通过对图像进行网格划分后分别计算多个局部单应性矩阵完成初步对准,然后对网格顶点进行微调优化完成最后配准。最后对图像进行融合形成无缝、自然的大视角图像,并利用场景信息在获取的全景图像上对目标进行快速定位,以满足监控人员对场景中目标的全景捕捉分析功能。实验结果表明,该算法能显著提高大场景下图像拼接结果的质量并实现目标的快速定位。

Local Homography Matrix Based Image Stitching and Location Algorithm

Due to the limited field of view of a single camera, the actual monitoring effect in large scenes is not ideal. An improved image stitching and target location algorithm is put forward for the problem. The algorithm is based on the surveillance images with overlapping regions acquired by multiple cameras. The initial alignment is completed by calculating local homography matrix after dividing the input images into uniform grids, and then the final alignment is achieved by fine-tuning the vertices of the grids. Finally, these images are fused to form a seamless and natural large-view image, and the scene information is used to rapidly locate the target in the panoramic image to satisfy the needs of monitoring and analyzing the target in the scene. The experimental results show that the proposed method can improve the quality of image stitching results effectively in large scenes and achieve fast target localization.