海洋信息技术与应用

针对鱼眼图像全景拼接中传统特征匹配算法效率低下，以及融合算法在畸变区域效果不佳的问题，本文提出了一种改进特征匹配与自适应融合的全景拼接算法。该方法基于多路鱼眼相机，通过棋盘格标定获取相机内参及畸变参数，并在相邻相机视场重叠区摆放标定板优化逆透视变换参数，从而建立相机间鸟瞰视图的映射关系。在此基础上，提出了一种基于原始图像畸变程度的自适应权重融合方法，动态调整融合权重，从而提升拼接质量。实验结果表明，该方法的拼接速度优于传统方法，有效消除了畸变区域的拼接缝，能够生成高质量的船艇全景俯视图像。

To address the inefficiency of traditional feature matching algorithms and the poor performance of fusion methods in distorted regions for fisheye image panoramic stitching, this paper proposes an improved panoramic stitching algorithm based on enhanced feature matching and adaptive fusion. The approach utilizes multiple fisheye cameras. The intrinsic and distortion parameters are obtained through chessboard calibration. Then calibration boards are placed in the overlapping fields of adjacent cameras to optimize inverse perspective transformation parameters and establish a mapping relationship between bird's-eye views of the cameras. An adaptive weighted fusion method, based on the distortion level of the original images, is introduced to dynamically adjust fusion weights, thereby enhancing the stitching quality. Experimental results demonstrate that the proposed method achieves higher stitching speed than traditional methods, effectively eliminates seams in distorted regions, and generates high-quality panoramic top-view images for ship assisted driving.