基于改进关键帧选取策略的快速PL-SLAM算法

doi:10.12263/DZXB.20210567

电子学报 ›› 2022, Vol. 50 ›› Issue (3): 608-618.DOI: 10.12263/DZXB.20210567

基于改进关键帧选取策略的快速PL-SLAM算法

陈孟元¹^,², 丁陵梅¹, 张玉坤¹

^1.安徽工程大学电气工程学院, 安徽芜湖 241000
^2.高端装备先进感知与智能控制教育部重点实验室，安徽芜湖 241000

收稿日期:2021-05-06 修回日期:2021-06-10 出版日期:2022-03-25
- 作者简介:
- 陈孟元　男,1984年1月生于安徽芜湖. 现为安徽工程大学电气工程学院副教授,硕士生导师.获安徽省科学技术奖一等奖. 主要研究方向为移动机器人SLAM、目标跟踪及路径规划.E-mail:mychen@ahpu.edu.cn
  丁陵梅　女,1994年12月出生于江苏省泰州市.现为安徽工程大学电气工程学院硕士研究生.研究方向为移动机器人视觉SLAM算法. E-mail:1425288603@qq.com
  张玉坤　男,1995年5出生于安徽省亳州市. 现为安徽工程大学电气工程学院硕士研究生. 研究方向为移动机器人视觉SLAM算法. E-mail:1728443478@qq.com
- 基金资助:
- 国家自然科学基金 (61903002); 芜湖市科技计划项目 (2020yf59); 安徽工程大学-鸠江区产业协同创新专项基金 (2021cyxtb8); 安徽工程大学中青年拔尖人才项目

Fast PL-SLAM Algorithm Based on Improved Keyframe Extraction Strategy

CHEN Meng-yuan¹^,², DING Ling-mei¹, ZHANG Yu-kun¹

^1.School of Electrical Engineering, Anhui Polytechnic University, Wuhu, Anhui 241000, China
^2.Key Laboratory of Advanced Perception and Intelligent Control of High?end Equipment, Wuhu, Anhui 241000, China

Received:2021-05-06 Revised:2021-06-10 Online:2022-03-25 Published:2022-03-25
- Supported by:
- National Natural Science Foundation of China (61903002); Wuhu Science and Technology Project of Anhui Province (2020yf59); Collaborative Innovation Fund of Anhui University of Technology and Jiujiang District Industries (2021cyxtb8); Young and Middle-aged Top-notch Talents Program of Anhui University of Technology

摘要/Abstract

摘要：

针对PL-SLAM（Point and Line Simultaneous Localization And Mapping）算法在稠密场景下同时使用点线特征造成特征计算冗余，以及曲线运动时漏选关键帧等问题，提出一种基于改进关键帧选取策略的快速PL-SLAM算法（Improved keyframe extraction strategy-based Fast PL-SLAM algorithm， IFPL-SLAM）.该算法引入一种基于信息熵引导的位姿跟踪决策，使用信息熵评价优先提取的特征点，依据评价结果决策点线特征的融合使用方式，避免了在纹理稠密场景下点线特征同时使用造成数据冗余，提高了算法的实时性;与此同时，为避免曲线运动时漏选关键帧，采用逆向索引关键帧选取策略补充在曲线运动中漏选的关键帧，提高了闭环的准确率和定位精度.在公开的KITTI数据集和TUM数据集中进行测试，测试结果表明本文算法的运行时间与PL-SLAM算法相比减少了16.0%，绝对轨迹误差相比于PL-SLAM算法缩小了23.4%，表现出了良好的构图能力.

关键词: 点线特征, 位姿估计, 信息熵, 关键帧, 逆向索引, 同时定位与地图构建

Abstract:

To address the problems that PL-SLAM(Point and Line Simultaneous Localization And Mapping) algorithm uses point and line features at the same time in dense scenes, which cause redundancy in feature calculation, and misses keyframe selection during curve motion, an IFPL-SLAM(Improved keyframe extraction strategy-based Fast PL-SLAM) algorithm is proposed. The algorithm introduces an information entropy-guided bit tracking decision, evaluates the priority extracted feature points using information entropy, and decides the fusion of point and line features based on the evaluation results, which avoids data redundancy caused by the simultaneous use of point and line features in dense texture scenes and improves the real-time performance of the algorithm. At the same time, in order to avoid missing keyframes during curve motion, the inverse index keyframe selection strategy is used to supplement the keyframes missed during curve motion, which improves the accuracy and positioning precision of the closed loop. The test results on the publicly available KITTI dataset and TUM dataset show that the running time of the algorithm proposed is reduced by 16.0% compared to the PL-SLAM algorithm, and the localization accuracy is increased by 23.4% compared to PL-SLAM, which exhibits a good mapping capability.

Key words: point and line features, position estimation, information entropy, keyframe, inverse index, simultaneous localization and mapping

中图分类号:

TP242.6

陈孟元, 丁陵梅, 张玉坤. 基于改进关键帧选取策略的快速PL-SLAM算法[J]. 电子学报, 2022, 50(3): 608-618.

Meng-yuan CHEN, Ling-mei DING, Yu-kun ZHANG . Fast PL-SLAM Algorithm Based on Improved Keyframe Extraction Strategy[J]. Acta Electronica Sinica, 2022, 50(3): 608-618.

图/表 14

参考文献 23

1	蒋林, 向超, 朱建阳, 刘奇. 加载语义似然估计的粒子滤波重定位[J]. 电子学报, 2021, 49(2): 306-314.
	JIANGLin, XIANGChao, ZHUJian-yang, LIUQi. Particle filter relocation with semantic likelihood estimation[J]. Acta Electronica Sinica, 2021, 49(2): 306-314. (in Chinese)
2	CHENW, LIUY. Research on unmanned system vision SLAM technology[C]//Proceedings of International Conference on Big Data, Artificial Intelligence and Internet of Things Engineering. Fuzhou, China: IEEE, 2020: 439-442.
3	陈宝华, 邓磊, 陈志祥, 段岳圻, 周杰. 基于即时稠密三维重构的无人机视觉定位[J]. 电子学报, 2017, 45(6): 1294-1300.
	CHENBao-hua, DENGLei, CHENZhi-xiang, DUANYue-qi, ZHOUJie. Instant dense 3D reconstruction-based UAV vision localization[J]. Acta Electronica Sinica, 2017, 45(6): 1294-1300. (in Chinese)
4	罗会兰, 陈鸿坤. 基于深度学习的目标检测研究综述[J]. 电子学报, 2020, 48(6): 1230-1239.
	LUOHui-lan, CHENHong-kun. Survey of object detection based on deep learning[J]. Acta Electronica Sinica, 2020, 48(6): 1230-1239. (in Chinese)
5	陆兴鹏, 王明羽, 曹扬, 赵瑞莲, 周武, 李兆麟, 魏少军. 一种基于帧图像的动态视觉传感器样本集建模方法[J]. 电子学报, 2020, 48(8): 1457-1464.
	LUXing-peng, WANGMing-yu, CAOYang, ZHAORui-lian, ZHOUWu, LIZhao-lin, WEIShao-jun. A dynamic vision sensor sample set modeling method based on frame images[J]. Acta Electronica Sinica, 2020, 48(8): 1457-1464. (in Chinese)
6	MUR-ARTALR, MONTIELJ M M, TARDOSJ D. ORB-SLAM: A versatile and accurate monocular SLAM system[J]. IEEE Transactions on Robotics, 2015, 31(5): 1147-1163.
7	ENGELJ, SCHÖPST, CREMERSD. LSD-SLAM: large-scale direct monocular SLAM[C]//Proceedings of European Conference on Computer Vision. Zurich, Switzerland: Springer, 2014: 834-849.
8	GOMEZ-OJEDAR, GONZALEZ-JIMENEZJ. Robust stereo visual odometry through a probabilistic combination of points and line segments[C]//Proceedings of IEEE International Conference on Robotics & Automation. Stockholm, Sweden: IEEE, 2016: 2521-2526.
9	NEIRAJ E, MAR I, RIBEIRO, et al. Mobile robot localization and map building using monocular vision[J]. Springer Berlin, 1997, 39(6): 275-284.
10	GIOI R GVON, JAKUBOWICZJ, J-MMOREL, RANDALLG. LSD: A fast line segment detector with a false detection control[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010, 32(4): 722-732.
11	ZHANGL, KOCHR. An efficient and robust line segment matching approach based on LBD descriptor and pairwise geometric consistency[J]. Journal of Visual Communication and Image Representation, 2013, 24(7): 794-805.
12	ZHANGG, JINH L, LIM J, et al. Building a 3-D line-based map using a stereo SLAM[J]. IEEE Transactions on Robotics, 2015, 31(6): 1-14.
13	ZUOX X, XIEX J, LIUY, et al. Robust visual SLAM with point and line features[C]//Proceedings of International Conference on Intelligent Robots and Systems. Piscataway, USA: IEEE, 2017: 1775-1782.
14	GOMEZ-OJEDAR, MORENOF, ZUÑIGA-NOËLD, SCARAMUZZAD, GONZALEZ-JIMENEZJ. PL-SLAM: a stereo SLAM system through the combination of points and line segments[J]. IEEE Transactions on Robotics, 2019, 35(3): 734-746.
15	MUR-ARTALR, TARD'OSJ D. ORB-SLAM2: An open-source SLAM system for monocular, stereo and RGB-D cameras[J]. IEEE Transactions on Robotics, 2017, 33(5): 1255-1262.
16	RUBLEEE, RABAUDV, KONOLIGEK, BRADSKIG. ORB: An efficient alternative to SIFT or SURF[C]//Proceedings of International Conference on Computer Vision. Barcelona, Spain: IEEE, 2011: 2564-2571.
17	HUANGS S, MAZ Y, MUT J, et al. Lidar-monocular visual odometry using point and line features[C]//Proceedings of International Conference on Robotics and Automation. Paris, France: IEEE, 2020: 1091-1097.
18	赵良玉, 金瑞, 朱叶青, 高凤杰. 基于点线特征融合的双目惯性SLAM算法[J].航空学报, 2021, 42(12): 325117.
	ZHAOLiang-yu, JINRui, ZHUYe-qing, GAOFeng-jie. Stereo visual-inertial slam with point and line features[J]. Acta Aeronautica et Astronautica Sinica, 2021,42(12): 325117. (in Chinese)
19	于雅楠, 卫红, 陈静. 基于局部熵的SLAM视觉里程计优化算法[J/OL]. 自动化学报. [2019-04-16]. .
	YUYa-nan, WEIHong, CHENJing. Optimization algorithm of visual odometry for SLAM based on local image entropy [J/OL]. Acta Automatica Sinica. [2019-04-16].(in Chinese)
20	郭小英, 李文书, 钱宇华, 白茹意, 贾春花. 可计算图像复杂度评价方法综述[J]. 电子学报, 2020, 48(4): 819-826.
	GUOXiao-ying, LIWen-shu, QIANYu-hua, BAIRu-yi, JIAChun-hua. Computational evaluation methods of visual complexity perception for images[J]. Acta Electronica Sinica, 2020, 48(4): 819-826. (in Chinese)
21	GALVEZ-LÓPEZD, TARDOSJ D. Bags of binary words for fast place recognition in image sequences[J]. IEEE Transactions on Robotics, 2012, 28(5): 1188-1197.
22	GEIGERA, LENZP, URTASUNR. Are we ready for autonomous driving? The KITTI vision benchmark suite[C]//Proceedings of Conference on Computer Vision and Pattern Recognition. Providence, RI, USA: IEEE, 2012: 3354-3361.
23	STURMJ, ENGELHARDN, ENDRESF, BURGARDW, CREMERSD. A benchmark for the evaluation of RGB-D SLAM systems[C]//Proceedings of International Conference on Intelligent Robots and Systems. Vilamoura-Algarve, Portugal: IEEE, 2012: 573-580.

TUM sequence	特征匹配准确率（%）			特征提取时间（ms）
TUM sequence	ORB-SLAM2	PL-SLAM	IFPL-SLAM	ORB-SLAM2	PL-SLAM	IFPL-SLAM
fr2/360-hemisphere	84.5	87.5	90.8	38.5	54.2	45.6
fr3/large/cabinet	76.5	85.7	89.4	40.6	57.3	50.1
fr1/plant	85.6	86.3	88.9	42.3	57.8	47.2
fr2/xyz	86.7	88.1	90.1	40.5	60.1	49.7
平均值	83.3	86.9	89.8	40.5	57.4	48.2

TUM sequence	特征匹配准确率（%）			特征提取时间（ms）
TUM sequence	ORB-SLAM2	PL-SLAM	IFPL-SLAM	ORB-SLAM2	PL-SLAM	IFPL-SLAM
fr2/360-hemisphere	84.5	87.5	90.8	38.5	54.2	45.6
fr3/large/cabinet	76.5	85.7	89.4	40.6	57.3	50.1
fr1/plant	85.6	86.3	88.9	42.3	57.8	47.2
fr2/xyz	86.7	88.1	90.1	40.5	60.1	49.7
平均值	83.3	86.9	89.8	40.5	57.4	48.2

KITTI sequence	提取的关键帧数量（帧）			单帧图像的处理时间（ms）			轨迹均方根误差RMSE/（m）
KITTI sequence	ORB- SLAM2	PL- SLAM	IFPL- SLAM	ORB- SLAM2	PL- SLAM	IFPL- SLAM	ORB- SLAM2	PL- SLAM	IFPL- SLAM
01	373	385	394	125	172	135	12.73	9.78	7.96
10	495	487	547	114	155	121	11.35	8.52	6.08
08	-	1784	1927	-	164	124	-	6.85	4.96
02	2097	2182	2285	134	181	156	9.27	7.89	6.31
平均值	988	1207	1288	124.33	168	134	11.12	8.26	6.33

KITTI sequence	提取的关键帧数量（帧）			单帧图像的处理时间（ms）			轨迹均方根误差RMSE/（m）
KITTI sequence	ORB- SLAM2	PL- SLAM	IFPL- SLAM	ORB- SLAM2	PL- SLAM	IFPL- SLAM	ORB- SLAM2	PL- SLAM	IFPL- SLAM
01	373	385	394	125	172	135	12.73	9.78	7.96
10	495	487	547	114	155	121	11.35	8.52	6.08
08	-	1784	1927	-	164	124	-	6.85	4.96
02	2097	2182	2285	134	181	156	9.27	7.89	6.31
平均值	988	1207	1288	124.33	168	134	11.12	8.26	6.33

TUM sequence	ORB-SLAM2		PL-SLAM		IFPL-SLAM
TUM sequence	平移误差（m）	旋转误差（°）	平移误差（m）	旋转误差（°）	平移误差（m）	旋转误（°）
fr3/large/cabinet	-	-	0.059	2.48	0.032	1.98
fr2/360-hemisphere	-	-	-	-	0.043	2.04
fr1/plant	0.092	4.52	0.054	3.12	0.048	2.95
fr2/xyz	0.038	2.24	0.029	1.83	0.023	1.24
平均值	0.065	3.38	0.047	2.48	0.037	2.053

基于改进关键帧选取策略的快速PL-SLAM算法

Fast PL-SLAM Algorithm Based on Improved Keyframe Extraction Strategy

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 14

参考文献 23

相关文章 15

编辑推荐

Metrics

[1]	王神龙, 雍宇, 吴晨睿. 基于伪孪生神经网络的低纹理工业零件6D位姿估计[J]. 电子学报, 2023, 51(1): 192-201.
[2]	张宏伟, 张小虎, 曹勇. 贝叶斯序贯重要性积分滤波器[J]. 电子学报, 2022, 50(4): 823-831.
[3]	胡钢, 牛琼, 许丽鹏, 卢志宇, 过秀成. 基于网络超链接信息熵的节点重要性序结构演化建模分析[J]. 电子学报, 2022, 50(11): 2638-2644.
[4]	魏钰轩, 陈莹. 基于自适应层信息熵的卷积神经网络压缩[J]. 电子学报, 2022, 50(10): 2398-2408.
[5]	李彦苍, 巩翔宇. 基于信息熵的改进狮群算法及其在组合优化中的应用[J]. 电子学报, 2021, 49(8): 1577-1585.
[6]	王树亮, 毕大平, 阮怀林, 杜明洋, 潘继飞. 基于信息熵准则的认知雷达机动目标跟踪算法[J]. 电子学报, 2019, 47(6): 1277-1284.
[7]	程玉胜, 赵大卫, 王一宾, 裴根生. 非平衡化标签补全核极限学习机多标签学习[J]. 电子学报, 2019, 47(3): 719-725.
[8]	吴宁博, 彭长根, 牟其林. 面向关联属性的差分隐私信息熵度量方法[J]. 电子学报, 2019, 47(11): 2337-2343.
[9]	冀中, 樊帅飞. 基于超图排序算法的视频摘要[J]. 电子学报, 2017, 45(5): 1035-1043.
[10]	邓大勇, 薛欢欢, 苗夺谦, 卢克文. 属性约简准则与约简信息损失的研究[J]. 电子学报, 2017, 45(2): 401-407.
[11]	李玉鑑, 张亚红. 三元变量间一维流形依赖关系的检测[J]. 电子学报, 2016, 44(3): 639-645.
[12]	孙雪, 李昆仑, 韩蕾, 白晓亮. 基于特征项分布的信息熵及特征动态加权概念漂移检测模型[J]. 电子学报, 2015, 43(7): 1356-1361.
[13]	谭丽, 孙季丰. 基于码书索引变换的高通量DNA序列数据压缩算法[J]. 电子学报, 2015, 43(5): 1007-1013.
[14]	唐振华, 梁祥严, 覃团发, 常侃. 分布式视频编码中基于多概率混合分布的相关噪声建模方法[J]. 电子学报, 2015, 43(2): 365-370.
[15]	吴成茂. 基于后验概率熵的正则化Otsu阈值法[J]. 电子学报, 2013, 41(12): 2474-2478.