为快速得到较大范围且清晰度足够的路面图像,提出一种针对航拍路面图像的拼接算法,可以快速准确处理图像,得到路面的拼接图像,为无人机航拍应用道路交通领域提供便利。路面图像拼接时,图像包含较大的相似区域,导致配准正确率较低,且图像信息量大,拼接时间长。为解决这一问题,本文提出在特征点提取前确定相邻图像的重叠区域,在重叠区域内使用尺度不变特征变换特征点结合稀疏表示匹配的方法,对相邻图像进行配准,随后利用小波变换方法,将图像进行融合,改善拼接效果. 经过实验证明,本文方法可以有效处理航拍路面图像,得到效果较好的路面拼接图像。

In order to quickly obtain a wide range of pavement images with sufficient clarity, a stitching algorithm for aerial pavement images is proposed, which can quickly and accurately process the images to obtain the stitched images of the pavement surface so as to provide convenience for the application of unmanned aerial vehicle aerial photography in the field of pavement transportation. When pavement images are stitched, the images contain large similar areas, resulting in a low registration accuracy, a large amount of image information, and a long stitching time. In order to solve this problem, this paper proposes to determine the overlapping area of adjacent images before feature point extraction. Scale-invariant feature transform feature points combined with sparse representation matching method in the overlapping area to register adjacent images. Then, wavelet transform method is used fuse the images to improve the stitching effect. Experimental results prove that the proposed method can effectively process aerial pavement images and obtain better pavement stitching images.