无人机影像物方多视匹配算法
发布时间：2018-07-14 10:42
【摘要】：目的像方无人机影像多视匹配方法忽视了影像之间的几何关系,而以MVLL(multi-view vertical line locus)为代表的物方多视匹配方法缺乏对地形之间相互约束的考虑。为此构建一种融合两类多视匹配方法优点的无人机影像物方多视匹配算法。方法在MVLL匹配结构的基础上添加半全局匹配的相容性约束,不仅继承了原半全局算法对有弱纹理区域匹配效果好和物体边缘突出的优点,而且摆脱了需制作核线影像的繁琐过程;采用物方窗口SNCC(summed normalized cross correlation)一致性匹配测度计算方法,有效降低摄影角度和遮挡对匹配结果的影响;采用金字塔分层的策略以提高匹配的速度和可靠性。结果选取自主研制的旋翼无人机三轴稳定平台获取了高分辨率无人机影像作为实验数据,从匹配效果、新匹配测度性能和匹配精度3个方面对算法进行了测试实验。本文算法整体匹配效果良好,物方窗口SNCC一致性匹配测度可有效消除匹配测度中的粗差,经过测定本文匹配算法生成的点云数据的高程精度为0.049 m,即约为1个GSD(ground space resolution)对应的地面大小。结论本文算法充分利用了无人机影像的多视信息进行匹配计算,具有匹配效果好、鲁棒性强和匹配精度高的优势。
[Abstract]:Objective the image multi-view matching method of image square UAV neglects the geometric relationship between images, while the object multi-view matching method represented by multi-view vertical line locus) lacks the consideration of the mutual constraint between terrain. In this paper, an object-square multi-view matching algorithm for UAV images is proposed, which combines the advantages of two kinds of multi-view matching methods. Based on the MVLL matching structure, the compatibility constraint of semi-global matching is added, which not only inherits the advantages of the original semi-global algorithm, but also has the advantages of good matching effect on weakly textured regions and prominent edges of objects. Moreover, it gets rid of the tedious process of making core line image, adopts the method of object window SNCC (summed normalized cross correlation) consistency matching measure, effectively reduces the influence of camera angle and occlusion on the matching result. Pyramid stratification strategy is adopted to improve the speed and reliability of matching. Results High-resolution UAV images were obtained from a self-developed three-axis stabilized platform for rotors. The algorithm was tested from three aspects: matching effect, performance of new matching measure and matching accuracy. The overall matching effect of this algorithm is good, and the coarse error in the matching measure can be effectively eliminated by the object window SNCC consistency matching measure. The height accuracy of the point cloud data generated by the matching algorithm in this paper is 0.049 m, that is, the ground size corresponding to about 1 GSD (ground space resolution). Conclusion the algorithm makes full use of the multi-view information of UAV images and has the advantages of good matching effect, strong robustness and high matching accuracy.
【作者单位】： 信息工程大学;地理信息工程国家重点实验室;航空遥感技术国家测绘地理信息局重点实验室;
【基金】：国家自然科学基金项目(41501482) 地理信息工程国家重点实验室开放研究基金项目(SKLGIE 2015-M-3-6和SKLGIE 2014-M-3-1) 航空遥感技术国家测绘地理信息局重点实验室开放基金项目(2014B02)~~
【分类号】：P231
本文编号：2121389
[Abstract]:Objective the image multi-view matching method of image square UAV neglects the geometric relationship between images, while the object multi-view matching method represented by multi-view vertical line locus) lacks the consideration of the mutual constraint between terrain. In this paper, an object-square multi-view matching algorithm for UAV images is proposed, which combines the advantages of two kinds of multi-view matching methods. Based on the MVLL matching structure, the compatibility constraint of semi-global matching is added, which not only inherits the advantages of the original semi-global algorithm, but also has the advantages of good matching effect on weakly textured regions and prominent edges of objects. Moreover, it gets rid of the tedious process of making core line image, adopts the method of object window SNCC (summed normalized cross correlation) consistency matching measure, effectively reduces the influence of camera angle and occlusion on the matching result. Pyramid stratification strategy is adopted to improve the speed and reliability of matching. Results High-resolution UAV images were obtained from a self-developed three-axis stabilized platform for rotors. The algorithm was tested from three aspects: matching effect, performance of new matching measure and matching accuracy. The overall matching effect of this algorithm is good, and the coarse error in the matching measure can be effectively eliminated by the object window SNCC consistency matching measure. The height accuracy of the point cloud data generated by the matching algorithm in this paper is 0.049 m, that is, the ground size corresponding to about 1 GSD (ground space resolution). Conclusion the algorithm makes full use of the multi-view information of UAV images and has the advantages of good matching effect, strong robustness and high matching accuracy.
【作者单位】： 信息工程大学;地理信息工程国家重点实验室;航空遥感技术国家测绘地理信息局重点实验室;
【基金】：国家自然科学基金项目(41501482) 地理信息工程国家重点实验室开放研究基金项目(SKLGIE 2015-M-3-6和SKLGIE 2014-M-3-1) 航空遥感技术国家测绘地理信息局重点实验室开放基金项目(2014B02)~~
【分类号】：P231
【相似文献】
相关期刊论文 前6条
1 韩心志;航天遥感多光谱扫描仪物方摆镜的高速摄影测试[J];高速摄影与光子学;1987年03期
2 游传金;Topocart-D型仪y对称性误差的分析与校正[J];测绘通报;1990年04期
3 张祖勋,张剑清,胡翔云;基于物方空间几何约束最小二乘匹配的建筑物半自动提取方法[J];武汉大学学报(信息科学版);2001年04期
4 王树根,李宾,王海涛,仲思东;基于车载CCD影像的物方几何参数量算[J];测绘信息与工程;2001年01期
5 侯文广;商浩亮;冯文灏;;二维控制场建立中的关键技术研究[J];测绘科学;2006年02期
6 ;[J];;年期
相关会议论文 前1条
1 王树根;李宾;王海涛;仲思东;;基于车载CCD影像的物方几何参数量算[A];新世纪 新机遇 新挑战——知识创新和高新技术产业发展（上册）[C];2001年
相关博士学位论文 前1条
1 樊仲谋;摄影测树原理与技术方法研究[D];北京林业大学;2015年
相关硕士学位论文 前1条
1 齐润冰;物方三角网约束的多视影像密集匹配[D];辽宁工程技术大学;2015年
，本文编号：2121389
本文链接：https://www.wllwen.com/kejilunwen/dizhicehuilunwen/2121389.html
