随着无人机摄影测量平台和计算机视觉技术的飞速发展。无人机摄影测量平台凭借便于携带、易操作、低成本、高效迅捷、机动性强等优点，在国土资源调查、三维建模、物体识别、目标追踪、测绘等领域中广泛应用，成为国内外众多学者研究的重点之一。由于无人机体积小、重量轻的原因，在飞行时易受到恶劣环境和风力的影响，容易产生旋转、视角偏差、平移等影像数据，对实现快速、高精度的影像匹配结果存在较大影响。因此本文基于传统特征匹配算法的原理与特性，对其进行改进，从而实现无人机影像高效率、高精度匹配。本文主要研究内容如下：

1、基于特征匹配算法，对常用的SIFT算法、SURF算法、ORB算法、KAZE算法、AKAZE算法的基本原理和匹配流程进行分析，采用Mikolajczyk数据集为实验数据，对5种算法进行特征匹配实验，从匹配速率和匹配准确率进行对比分析，对比各算法的鲁棒性。

2、针对基于特征匹配算法是以灰度图像为输入对象进行影像匹配，对无人机影像的色彩特征信息没有进行充分利用，所以本文用AKAZE算法进行影像特征检测特征，利用具有颜色空间信息的Opponent-DAISY描述子进行特征描述，使用3组无人机影像对其进行不同变量（旋转、尺寸、亮度、模糊）变化，对不同变量变化的影像进行特征匹配实验，验证改进算法的鲁棒性，并于与传统算法进行比较。

3、针对RANSAC算法进行精确匹配时其迭代次数没有限制，人为进行设限其并不一定最优结果点。引入VFC算法对无人机影像进行精确匹配，在色彩对比度弱的环境下匹配效果差等问题。本文采用K-VFC相结合的算法进行误匹配点的剔除，以此来提高无人机影像匹配的准确度。通过实验发现，K-VFC算法的匹配准确度可达到80%以上，相较于RANSAC算法和VFC算法准确度提高了15%，表明K-VFC算法可以有效地提高无人机影像匹配的效果。

With the rapid development of UAV photogrammetry platform and computer vision technology. UAV photogrammetry platforms are widely used in land and resources surveys, three-dimensional modeling, object recognition, target tracking, surveying and mapping and other fields due to their advantages such as portability, easy operation, low cost, high efficiency and quickness, and strong maneuverability. One of the focuses of many scholars' research. Due to the small size and light weight of the UAV, it is easily affected by harsh environments and wind during flight, and it is easy to generate image data such as rotation, viewing angle deviation, and translation, which has a great impact on the realization of fast and high-precision image matching results. . Therefore, based on the principles and characteristics of the traditional feature matching algorithm, this paper improves it to achieve high-efficiency and high-precision matching of UAV images. The main research contents of this paper are as follows:
1. Based on the feature matching algorithm, analyze the basic principles and matching process of the commonly used SIFT algorithm, SURF algorithm, ORB algorithm, KAZE algorithm, and AKAZE algorithm. Using the Mikolajczyk data set as the experimental data, perform feature matching experiments on the five algorithms. Compare and analyze the matching rate and matching accuracy rate, and compare the robustness of each algorithm.
2. Aiming at the image matching based on the feature matching algorithm that uses the gray image as the input object, the color feature information of the UAV image is not fully utilized, so this paper uses the AKAZE algorithm to perform the image feature detection feature, and uses the color space information The Opponent-DAISY descriptor is used to describe features, using 3 sets of UAV images to change different variables (rotation, size, brightness, blur), and perform feature matching experiments on images with different variable changes to verify the robustness of the improved algorithm , And compare it with traditional algorithms.
3. There is no limit to the number of iterations of the RANSAC algorithm for precise matching, and it is not necessarily the optimal result point to artificially set the limit. The VFC algorithm is introduced to accurately match UAV images, and the matching effect is poor in an environment with weak color contrast. In this paper, the K-VFC combined algorithm is used to eliminate mismatch points, so as to improve the accuracy of UAV image matching. Through experiments, it is found that the matching accuracy of the K-VFC algorithm can reach more than 80%, which is 15% higher than the accuracy of the RANSAC algorithm and the VFC algorithm, indicating that the K-VFC algorithm can effectively improve the effect of UAV image matching.

[1]朱述龙. 遥感图像处理与应用[M]. 科学出版社, 2006.

[2]王静. 面向国土资源巡查的无人机影像拼接算法研究[D]. 中国矿业大学,2019.

[3]杜蒙蒙,刘颖超,姬江涛,等. 基于无人机与激光测距技术的农田地形测绘[J]. 农业工程学报,2020, 36(22): 60-67.

[4]林勇,李旭涛,吴崧源,等. 固定翼无人机优于2cm航测在地形测量中的应用[J]. 测绘科学, 2019, 44(10): 196-204.

[5]张海艳,兰玉彬,文晟,等. 植保无人机旋翼风场模型与雾滴运动机理研究进展[J]. 农业工程学报,2020,36(22):1-12.

[6]Lan Y B,Chen S D,Fritz B K.Current status and future trends of precision agricultural aviation technologies[J].Int JAgric Biol Eng,2017,10(3):1-17.

[7]张东彦,兰玉彬,陈立平,等. 中国农业航空施药技术研究进展与展望[J]. 农业机械学报，2014,45(10):53-59.

[8]宋伟. 基于改进AKAZE特征的无人机影像匹配算法研究[D]. 中国矿业大学,2019.

[9]朱俊锋. 利用倾斜影像进行三维数字城市重建的关键技术研究[D]. 武汉大学,2014

[10]张一.无人机遥感影像点特征匹配算法研究[D]. 解放军信息工程大学, 2015.

[11]贾银江. 无人机遥感图像拼接关键技术研究[D]. 东北农业大学, 2016.

[12]唐敏. 序列无人机影像预处理与匹配中的几个关键问题研究[D]. 西南交通大学, 2014.

[13]秦博,王蕾. 无人机发展综述[J]. 飞航导弹, 2002(08):4-10.

[14]王新,陈武,汪荣胜,等. 浅论低空无人机遥感技术在水利相关领域中的应用前景[J]. 浙江水利科技, 2010(06):27-29.

[15]吴洁琼. 基于特征的图像匹配算法研究与应用[D]. 2016.

[16]Ma W, Di Q, Xiong Z, et al. A Quantitative Assessment Method of Image Matching Methods Based on Cumulative Prospect Theory[C]// International Conference on Intelligent Human-machine Systems & Cybernetics. 2016.

[17]徐晓伟. 无人机影像匹配算法研究[D]. 2019.

[18]Dubuisson M P, Jain A K. A modified Hausdorff distance for object matching[C]// Pattern Recognition, 1994. Vol. 1 - Conference A: Computer Vision & Image Processing. Proceedings of the 12th IAPR International Conference on. IEEE, 1994.

[19]Barnea D I, Silverman H F. A Class of Algorithms for Fast Digital Image Registration[J]. IEEE Transactions on Computers, 1972, C-21(2):179-186.

[20]Hanebeck U D. Template matching using fast normalized cross correlation[J].Proceeding of Spie on Optical Pattern Recognition XII, 2001, 4387:95-102.

[21]郑军,何叶松,谢国明,等. 基于划分强度一致测度的多模医学图像配准[J]. 中国医科大学学报, 2006(01):94-95.

[22]吴培景,陈光梦. 一种改进的SSDA图像匹配算法[J]. 计算机工程与应用, 2005(33):80-82.

[23]王振江,吴健,林方全. 一种结合快速灰度投影与SSDA的图像匹配方法[J]. 计算机工程与应用, 2011,47(33):195-197.

[24]兰海滨,王平,赵保军. 基于色调空间的彩色图像匹配算法[J]. 计算机应用研究, 2009,26(03):1174-1176+1188.

[25]谢维达,周宇恒,寇若岚. 一种改进的快速归一化互相关算法[J]. 同济大学学报(自然科学版), 2011,39(08):1233-1237.

[26]吴鹏,徐洪玲,宋文龙. 结合小波金字塔的快速NCC图像匹配算法[J]. 哈尔滨工程大学学报, 2017,38(05):791-796.

[27]朱明,姚强,唐俊,等. 超图约束和改进归一化互相关方法相结合的图像配准算法[J].国防科技大学学报, 2019, 41(03):50-55.

[28]杨建文. 基于改进SIFT算法的倾斜影像匹配研究[D]. 中国矿业大学, 2019.

[29]钟用. 基于点特征的遥感图像配准[D].西安电子科技大学, 2018..

[30]陈凯. 基于特征的遥感图像配准方法研究[D].西安电子科技大学, 2014.

[31]Moravec H P. Towards Automatic Visual Obstacle Avoidance[C]//Proceedings of the 5th international joint conference on Artificial intelligence - Volume 2. Morgan Kaufmann Publishers Inc. 1977.

[32]Morevec H P. Towards automatic visual obstacle avoidance[C]. international joint conference on artificial intelligence, 1977: 584-584.

[33]C Harris, M Stephens . A combined corner and edge detector[J]. In Proc. of Fourth Alvey Vision.

[34]Lowe D G . Distinctive Image Features from Scale-Invariant Keypoints[J]. International ournal of Computer Vision, 2004, 60(2):91-110.

[35]Bay H , Tuytelaars T , Gool L V . SURF: Speeded up robust features[J]. 2006.

[36]Rosten E , Porter R , Drummond T . Faster and better: a machine learning approach to corner detection[J]. IEEE Trans Pattern Anal Mach Intell, 2008, 32(1):105-119.

[37]Rublee E , Rabaud V , Konolige K , et al. ORB: An efficient alternative to SIFT or SURF[J]. 2011.

[38]Lcantrilia P, Nuevo J, Bartoli A. Fast Explicit Diffusion for Accelerated Featuresin Nonlinear Scale Spaces: British Machine Vision Conference, 2013[C].

[39]Weuckert J, Grewenig S, Schroers C, et al. Cyclic Schemes for PDE-Based Image Analysis[J]. International Journal of Computer Vision, 2016,118(3):275-299.

[40]Grewening S, WEICKERT J, BRUHN A. From Box Filtering to Fast Explicit Diffusion[M]. Springer Berlin Heidelberg, 2010: 533-542.

[41]韩宇,宗群,邢娜. 基于改进SIFT的无人机航拍图像快速匹配[J]. 南开大学学报(自然科学版), 2019,52(01):5-9.

[42]刘利锋,马燕,张相芬,等. 采用二值SIFT特征描述的图像匹配方法[J]. 计算机应用与软件, 2016,33(12):152-155+210.

[43]姚永祥,李佳,段平,等. 联合对数极坐标描述与位置尺度特征的无人机影像匹配[J]. 武汉大学学报(信息科学版):1-11[2021-04-11].

[44]徐瑞瑞,雷添杰,程结海,等. 一种改进SURF算法的无人机影像快速配准方法[J]. 水利水电技术: 1-12[2021-04-11].

[45]李闯,王晓红,何志伟. 利用A-AKAZE算法进行喀斯特地区无人机影像匹配[J]. 测绘通报, 2020(06):53-56.

[46]张宇,胡立坤,黄彬. 融合颜色和光照信息的改进AKAZE图像配准算法[J]. 计算机工程:1-11[2021-04-11].

[47]Calonder M, Lepetit V, Strecha C, et al. BRIEF: Binary robust independent elementary features[C].European Conference on Computer Vision, 2010: 778-792.

[48]Tola E, Lepetit V, Fua P. DAISY: An Efficient Dense Descriptor Applied to Wide-Baseline Stereo[J]. IEEE Transactions on Software Engineering, 2010, 32(5):815-830.

[49]Vandergheynst P, Ortiz R, Alahi A. FREAK: Fast Retina Keypoint[C]// IEEE Conference on Computer Vision & Pattern Recognition. 2012.

[50]Yi K M, Trulls E, Lepetit V, et al. LIFT: Learned Invariant Feature Transform[C]//European Conference on Computer Vision. 2016.

[51]Xu X,Tian L,Feng J,Zhou J. OSRI: a rotationally invariant binary descriptor.[J]. IEEE transactions on image processing : a publication of the IEEE Signal Processing Society, 2014, 23(7).

[52]刘红敏,李璐,王志衡. 基于采样点组二值化策略的鲁棒二值描述子研究[J]. 计算机科学, 2017,44(12):292-297+309.

[53]张小宇. 基于低空无人机影像特征的匹配策略研究[D]. 西安科技大学, 2019.

[54]Fischler M A, Bolles R C. Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography[J]. Communications of the Acm, 1981, 24(6):381-395.

[55]Bian J W, Lin W Y, Matsushita Y, et al. Gms: Grid-based motion statistics for fast, ultra-robust feature correspondence[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017: 4181-4190.

[56]崔志祥,蓝朝桢,熊新,等. 一种无人机热红外与可见光影像匹配方法[J]. 测绘科学技术学报, 2019, 36(06):609-613.

[57]孙灏,高俊强,许苏苏. 基于SURF算法和改进RANSAC算法的无人机影像匹配[J]. 测绘工程, 2017, 26(11):55-59+64

[58]张霞,汤森林. 一种去除遥感图像条带噪声的小波矩匹配方法[J]. 遥感技术与应用, 2018, 33(02):305-312.

[59]Kuglin C D. The phase correlation image alignment method[J]. Proc.intl Conf.cybernetics & Society, 1975:163-165.

[60]DE C E, Morandi C. Registration of translated and rotated images using finite fourier transforms[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 1987, PAMI-9(5):700-703.

[61]何梦梦,郭擎,李安,等. 特征级高分辨率遥感图像快速自动配准[J]. 遥感学报, 2018, 22(02):277-292.

[62]胡梦云. 基于SURF和KAZE的图像配准算法研究[D]. 江西理工大学, 2016.

[63]沈元. 基于改进的SIFT算法图像匹配的研究[D]. 河北工业大学, 2014.

[64]陈鑫彤. 基于非线性金字塔的图像局部特征匹配研究[D]. 辽宁工程技术大学, 2019.

[65]姚立平,潘中良. 基于改进的HOG和LBP算法的人脸识别方法研究[J]. 光电子技术, 2020, 40(02):114-118+124.

[66]郭仙草,石美红,李青. 基于改进WLD的纹理特征提取方法[J]. 计算机工程, 2015, 41(04):210-216.

[67]闫蓓蓓,娄小平. DAISY和LBP描述符结合的三维点云粗拼接方法[J]. 电子测量与仪器学报, 2020, 34(08):73-78.

[68]查易艺,孙权森,罗楠,等. 基于FAST和DAISY的遥感图像配准算法[J]. 计算机应用研究, 2016, 33(02):624-628.

[69]傲天慈. 基于改进SURF算法的图像匹配[D]. 大连理工大学, 2019.

[70]朱军桃,龚朝飞,赵苗兴,等. 基于ORB算法和改进KNN-RANSAC算法的无人机遥感影像拼接[J]. 桂林理工大学学报, 2020, 40(01):131-137.

[71]林涛,赵璨. 最近邻优化的k-means聚类算法[J]. 计算机科学, 2019, 46(S2):216-219.