摘要
基于特征点检测的高分辨率图像拼接运算量大、速度慢,改进后可以提高效率。从减少参与SIFT计算的数据量和优化算法两方面实现图像拼接加速。建立基于ED-SIFT的图像拼接算法,图像边缘检测优选Soble算子;使用Python+OpenCV,对WorldView3切片数据进行拼接实验,平均耗时为0.96 s;与传统基于SIFT的图像拼接算法相比,在保证一定数目的特征点同时,速度加快9.47倍。使用该算法对重叠率为20%、65%、80%的切片数据拼接成功,平均耗时都在0.96~1.08 s。继续用OpenCV+Numba并行运算实现该算法,对运算提速已不明显。基于ED-SIFT的图像拼接算法对高分辨率遥感图像拼接提速明显,对不同重叠率图像拼接皆有效。
The high-resolution image mosaic based on feature point detection has a large amount of computation and slow speed, so it needs to be optimized to improve the efficiency.The image mosaic can be acceleratted by reducing the amount of data involved in SIFT calculation of optimizing the algorithm.The algorithm of image mosaic based on ED-SIFT was established based on Sobel kernel optimization.The mosaic experiment of worldview 3 slice data with 1210 × 630 size was carried out based on Python+OpenCV,and the time was 0.96s.Compared with the traditional image mosaic algorithm based on SIFT,it was about 9.47 times faster and has certain numbers of characteristic points.Using this algorithm, images of 20%,65%,80% overlapped slice data can be mosaiced successful, and the time consumption is within 0.96s to 1.08s.The Opencv + Numba parallel operation is used to realize this algorithm, and the speed of operation is not obvious.The image mosaic algorithm based on ED-SIFT can improve the speed of high-resolution remote sensing image mosaic obviously, and it is effective for image mosaic with different overlapping rate.
作者
杨云源
陈瑞
YANG Yunyuan;CHEN Rui(School of Resources,Environment and Chemistry,Chuxiong Normal University,Chuxiong 675000,China;School of Management and Economics,Chuxiong Normal University,Chuxiong 675000,China)
出处
《测绘工程》
2023年第1期8-13,共6页
Engineering of Surveying and Mapping
基金
云南省教育厅基金资助项目(2017ZZX018)
楚雄师范学院教改项目(19004)。