基于深度学习的无人机航测三维重建算法优化

中国民航大学学报 ›› 2024, Vol. 42 ›› Issue (6): 82-90.

基于深度学习的无人机航测三维重建算法优化

杨永刚 1 ，李思萌 2

1. 中国民航大学交通科学与工程学院，天津

300300； 2. 中国信息安全研究院有限公司智能网络所，北京

102200

收稿日期:2023-04-14 修回日期:2023-06-20 出版日期:2025-04-08 发布日期:2025-04-08
作者简介:杨永刚（1978— ），男，吉林四平人，教授，博士，研究方向为无人机任务规划等

Optimization of 3D reconstruction algorithm for UAV aerial survey based on

deep learning

YANG Yonggang 1 , LI Simeng 2

1. College of Transportation Science and Engineering, CAUC, Tianjin 300300, China;

2. Intelligent Network Institute, China Information Security Research Institute Co., Ltd., Beijing 102200, China

Received:2023-04-14 Revised:2023-06-20 Online:2025-04-08 Published:2025-04-08

摘要/Abstract

摘要：

利用传统的运动恢复结构（SFM, struct from motion）算法实现无人机视角下的三维重建时，为减少特征点的

错误匹配和运动目标对整体稀疏点云的影响，主要依靠随机采样一致性（Ransac, random sample consensus）算法，但会导致 Ransac 在求解相机位姿时的正确率降低，迭代次数增加。本文基于深度学习的单次多框

检测器（SSD，single shot multibox detector）网络进行目标检测。首先，在尺度不变特征转换（SIFT，scale-invariant feature transform）提取特征点后，对动态目标类别范围内的特征点进行剔除；然后，在 K 近邻（KNN，

K-nearest neighbor）暴力匹配后，对错误匹配进行剔除，减少无效运动目标范围内的特征点和不同类别间的

错误匹配，使得相同置信度时， Ransac 求解相机位姿的迭代次数降低，同时减少特征点暴力匹配和 SFM 算

法计算三维点的时间。最后，通过 2 个场景 12 张图片验证了引入深度学习优化后三维重建算法的可行性。

关键词:

无人机（UAV）, 三维重建, 目标检测, 单次多框检测器（SSD）

Abstract:

When the traditional struct from motion (SFM) algorithm is used to realize 3D reconstruction from the perspective of

unmanned aerial vehicle (UAV), in order to reduce the mismatching of feature points and the impact of moving targets on the overall sparse point cloud, the random sample consensus (Ransac) algorithm is mainly relied on. However, these problems can lead to a decrease in the accuracy and an increase in the number of iterations of Ransac

when solving camera poses. This article conducts target detection based on a deep learning single shot multibox detector (SSD) network. Firstly, feature points within the range of dynamic target categories are removed after scaleinvariant feature transform (SIFT) extraction of feature points. Then, mismatches are removed after K-nearest neighbor (KNN) violent matching to reduce feature points within the range of invalid moving targets and mismatching

between different categories. So that when the confidence is the same, the number of iterations of Ransac when

solving camera pose is reduced, and the time of feature point violence matching and SFM algorithm calculation of

3D points are also reduced. Finally, the feasibility of the 3D reconstruction algorithm optimized by deep learning

was verified through 12 images of 2 scenes.

Key words:

unmanned aerial vehicle (UAV), 3D reconstruction, target detection, single shot mutibox detector (SSD)

中图分类号:

V279+ .2

杨永刚, 李思萌.

基于深度学习的无人机航测三维重建算法优化

[J]. 中国民航大学学报, 2024, 42(6): 82-90.

YANG Yonggang , LI Simeng.

Optimization of 3D reconstruction algorithm for UAV aerial survey based on

deep learning

[J]. Journal of Civil Aviation University of China, 2024, 42(6): 82-90.

参考文献 18

[1]	包晓敏, 白晨．单目三维重建技术综述[J]．智能计算机与应用,
	2020, 10(7): 49-52．
[2]	张彦雯, 胡凯, 王鹏盛．三维重建算法研究综述[J]．南京信息工程
	大学学报(自然科学版), 2020, 12(5): 591-602．
[3]	李文豪．三维地图重建及充分利用无人机的探讨[J]．电子世界,
	2020(11): 56-57．
[4]	赵烨, 蒋建国, 洪日昌．基于 RANSAC 的 SIFT 匹配优化[J]．光电工
	程, 2014, 41(8): 58-65．
	[5]
	FISCHLER M, BOLLES R． Random sample consensus: a paradigm for
	model fitting with applications to image analysis and automated cartog－
	raphy[J]． Communications of the ACM, 1981, 24(6): 381-395．
	[6]
	WEI T, LIU H M, DONG Z L, et al． Robust monocular SLAM in dynam－
	ic environments[C]//2013 IEEE International Symposium on Mixed and
	Augmented Reality (ISMAR), October 1-4, 2013, Adelaide, SA, Australi－
	a． IEEE, 2013: 209-218．
	[7]
	SUN Y X, LIU M, MENG M Q H． Improving RGB-D SLAM in dynamic
	environments: a motion removal approach[J]． Robotics and Autonomous
	Systems, 2017, 89: 110-122．
[8]	BESCOS B, FACIL J M, CIVERA J, et al． DynaSLAM: tracking, mapping,
	and inpainting in dynamic scenes[J]． IEEE Robotics and Automation Letters, 2018, 3(4): 4076-4083．
[9]	佟帅, 徐晓刚, 易成涛, 等．基于视觉的三维重建技术综述[J]．计算
	机应用研究, 2011, 28(7): 2411-2417．
[10]	吴文琪, 孙增圻．机器视觉中的摄像机定标方法综述[J]．计算机应用
	研究, 2004, 21(2): 4-6．
[11]	ZHANG Z Y. Flexible camera calibration by viewing a plane from un－
	known orientations[C]//The 7th IEEE International Conference on Computer Vision, September 20-27, 1999, Kerkyra, Greece. IEEE, 1999:
	666-673.
[12]	ZHANG Z． A flexible new technique for camera calibration [J]． IEEE
	Transactions on Pattern Analysis and Machine Intelligence, 2000, 22
(11)	: 1330-1334．
[13]	BROWN M, LOWE D G． Automatic panoramic image stitching using in－
	variant features[J]． International Journal of Computer Vision, 2007, 74
(1)	: 59-73．
[14]	HARRIS C, STEPHENS M. A combined corner and edge detector[C]//
	Alvey Vision Conference 1988, Manchester． Alvey Vision Club, 1988:
	147-151．
[15]	王崴, 唐一平, 任娟莉, 等．一种改进的 Harris 角点提取算法[J]．
	光学精密工程, 2008, 16(10): 1995-2001．
[16]	陈科圻, 朱志亮, 邓小明, 等．多尺度目标检测的深度学习研究综述
[J]	．软件学报, 2021, 32(4): 1201-1227．
[17]	LIU W, ANGUELOV D, ERHAN D, et al． SSD: single shot MultiBox de－
	tector[C]//European Conference on Computer Vision． Cham: Springer,
	2016: 21-37．
[18]	李双杰, 张开翔, 王士栋, 等．基于加权 K 近邻的特征选择方法[J]．
	天津师范大学学报(自然科学版), 2020, 40(2): 63-67．