基于自适应深度约束的水下双目图像特征匹配

田嘉禾; 王宁; 陈廷凯; 李春艳; 陈帅

doi:10.19693/j.issn.1673-3185.02197

基于自适应深度约束的水下双目图像特征匹配

doi: 10.19693/j.issn.1673-3185.02197

田嘉禾^1,,
王宁^{2, 3, ,},
陈廷凯^1,,
李春艳¹,
陈帅¹

1.
大连海事大学船舶电气工程学院，辽宁大连 116026
2.
大连海事大学轮机工程学院，辽宁大连 116026
3.
哈尔滨工程大学船舶工程学院，黑龙江哈尔滨 150001

基金项目: 辽宁省“兴辽英才计划”资助项目（XLYC1807013）；装备预研重点实验室基金资助项目（6142215200106）；基础加强计划重点基础研究项目资助（2020-JCJQ-ZD-153）；青年科学家培育基金资助项目（79000006）

详细信息

作者简介:
田嘉禾，男，1996年生，硕士。研究方向：双目视觉三维重建。E-mail：tianj_h@163.com

王宁，男，1983年生，博士，教授，博士生导师。研究方向：无人系统建模与控制，自适应智能控制，机器学习，非线性控制。E-mail：n.wang.dmu.cn@gmail.com

陈廷凯，男，1993年生，博士生。研究方向：目标识别与检测。E-mail：tingkai.chen@aliyun.com

通信作者:
王宁

中图分类号: U662.9
计量
- 文章访问数: 176
- HTML全文浏览量: 89
- PDF下载量: 23
- 被引次数: 0
出版历程
- 收稿日期: 2020-11-23
- 修回日期: 2021-03-26
- 网络出版日期: 2021-11-09
- 刊出日期: 2021-12-20

Adaptive depth constraint-based underwater binocular image feature matching

Jiahe TIAN^1
,,
Ning WANG^{2, 3
, ,},
Tingkai CHEN^1
,,
Chunyan LI¹,
Shuai CHEN¹

1.
College of Marine Electrical Engineering, Dalian Maritime University, Dalian 116026, China
2.
College of Marine Engineering, Dalian Maritime University, Dalian 116026, China
3.
College of Shipbuilding Engineering, Harbin Engineering University, Harbin 150001, China

基于自适应深度约束的水下双目图像特征匹配由田嘉禾，等创作，采用知识共享署名4.0国际许可协议进行许可。

摘要

摘要: 目的针对水下双目图像特征点稀疏、极线约束模型失效等难题，提出一种基于自适应深度约束的水下图像特征匹配(ADC-UFM)算法。方法结合FAST算子与SIFT描述子，提高图像匹配精度；提出基于水下折射因子的特征匹配约束模型（MCM），有效剔除误匹配点；提出自适应阈值选取（ATC）方法，最大限度地保留复杂水下环境下的图像特征信息。结果实验结果显示，ADC-UFM算法优于现有的SIFT，SURF和UCC-SIFT等典型方法，匹配准确率可达85.2%，满足实时匹配需求。结论研究成果可为基于双目视觉系统的水下三维重建提供关键技术支撑。
- 自适应深度约束 /
- 水下双目图像 /
- 特征匹配 /
- 误匹配点剔除
Abstract: Objective In this paper, to address sparse feature points and unique epipolar constraints, an adaptive depth constraint-based underwater feature matching (ADC-UFM) scheme is proposed. Methods By combining a features from accelerated segment test (FAST) operator with scale invariant feature transform (SIFT) descriptors, the matching accuracy can be dramatically improved. By introducing an underwater refractive factor, the matching constraint model (MCM) can be effectively established, thereby contributing to eliminating mismatched points. The adaptive threshold choosing (ATC) module is finely devised to preserve image feature information in changeable underwater environments to an extreme extent. Results Comprehensive experiments show that the proposed ADC-UFM scheme can outperform typical matching schemes including SIFT, speeded-up robust features (SURF) and SIFT feature matching based on underwater curve constraint (UCC-SIFT), which not only achieves 85.2% matching accuracy but also meets the real-time requirements. Conclusion The results of this study can provide a reliable guarantee for subsequent underwater 3D reconstruction based on the binocular vision system.
- adaptive depth constraint (ADC) /
- underwater binocular images /
- feature matching /
- mismatched point elimination

HTML全文

图 1 FAST特征检测原理图

Figure 1. Schematic diagram of FAST feature detection

下载: 全尺寸图片幻灯片

图 2 关键点主方向分配原理图

Figure 2. Schematic diagram of main direction assignment of key points

下载: 全尺寸图片幻灯片

图 3 水下反投影模型

Figure 3. Underwater back projection model

下载: 全尺寸图片幻灯片

图 4 水下图像匹配流程图

Figure 4. Flow chart of underwater image matching

下载: 全尺寸图片幻灯片

图 5 水下数据采集平台

Figure 5. Underwater data acquisition platform

下载: 全尺寸图片幻灯片

图 6 不同数据集匹配结果

Figure 6. Matching results under different data sets

下载: 全尺寸图片幻灯片

图 7 不同算法匹配准确率对比图

Figure 7. Comparison of matching accuracy of different algorithms

下载: 全尺寸图片幻灯片

表 1 双目相机内部参数

Table 1. Internal parameters of binocular camera

参数	左相机	右相机
(f_x, f_y)	(1 201.30, 1 200.86)	(1 211.81, 1 207.01)
(u₀, v₀)	(319.24, 255.25)	(322.34, 316.32)
K	(0.28, −0.74, 0.01, 0.03, 0.0)	(−0.10, −0.06, 0.02, 0.01, 0.00)

下载: 导出CSV

表 2 实验结果数据对比

Table 2. Comparison of experimental data

数据来源		算法	匹配对数	正确率/%	匹配时间/s
采集的数据	暗光	SIFT	53	84.9	0.108
		SURF	64	84.3	0.060
		UCC-SIFT	26	84.6	1.930
		ADC-UFM	52	86.5	0.721
	旋转	SIFT	98	82.7	0.112
		SURF	103	83.4	0.062
		UCC-SIFT	52	84.6	1.963
		ADC-UFM	73	84.9	0.730
	浑浊	SIFT	34	79.4	0.093
		SURF	70	78.5	0.058
		UCC-SIFT	29	86.2	1.884
		ADC-UFM	40	82.5	0.710
公开数据	暗光	SIFT	62	81.3	0.098
		SURF	80	82.5	0.060
		UCC-SIFT	43	83.7	1.901
		ADC-UFM	60	85.0	0.712
	旋转	SIFT	121	84.2	0.293
		SURF	107	83.1	0.061
		UCC-SIFT	62	85.4	1.970
		ADC-UFM	90	85.6	0.732
	浑浊	SIFT	172	84.3	0.300
		SURF	195	83.5	0.064
		UCC-SIFT	98	85.7	1.991
		ADC-UFM	120	86.7	0.753

下载: 导出CSV

参考文献(19)

[1]	KAWAI R, YAMASHITA A, KANEKO T. Three-dimensional measurement of objects in water by using space encoding method[C]//Proceedings of 2009 IEEE International Conference on Robotics and Automation. Kobe, Japan: IEEE, 2009: 2830−2835.
[2]	BROWN M Z, BURSCHKA D, HAGER G D. Advances in computational stereo[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2003, 25(8): 993–1008. doi: 10.1109/TPAMI.2003.1217603
[3]	LOWE D G. Distinctive image features from scale-invariant keypoint[J]. International Journal of Computer Vision, 2004, 60(2): 91–110. doi: 10.1023/B:VISI.0000029664.99615.94
[4]	BAY H, ESS A, TUYTELAARS T, et al. Speeded-up robust features (SURF)[J]. Computer Vision and Image Understanding, 2008, 110(3): 346–359. doi: 10.1016/j.cviu.2007.09.014
[5]	KE Y, SUKTHANKAR R. PCA-SIFT: a more distinctive representation for local image descriptors[C]//Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR). Washington, DC, USA: IEEE, 2004: 506–513.
[6]	RUBLEE E, RABAUD V, KONOLIGE K, et al. ORB: an efficient alternative to SIFT or SURF[C]//Proceedings of 2011 International Conference on Computer Vision (ICCV). Barcelona, Spain: IEEE, 2011: 2564–2571.
[7]	NEGAHDARIPOUR S, SARAFRAZ A. Improved stereo matching in scattering media by incorporating a backscatter cue[J]. IEEE Transactions on Image Processing, 2014, 23(12): 5743–5755. doi: 10.1109/TIP.2014.2358882
[8]	李雅倩, 张岩松, 李海滨, 等. 基于深度约束的水下稠密立体匹配[J]. 光子学报, 2017, 46(7): 0715001-1−0715001-10. doi: 10.3788/gzxb20174607.0715001 LI Y Q, ZHANG Y S, LI H B, et al. Underwater dense stereo matching based on depth constraint[J]. Acta Photonica Sinica, 2017, 46(7): 0715001-1−0715001-10 (in Chinese). doi: 10.3788/gzxb20174607.0715001
[9]	GEDGE J. Underwater stereo matching and its calibration[D]. Edmonton: University of Alberta, 2011.
[10]	FERREIRA R, COSTEIRA J P, SANTOS J A. Stereo reconstruction of a submerged scene[M]//MARQUES J S, PÉREZ DE LA BLANCA N, PINA P. Pattern Recognition and Image Analysis. Berlin, Heidelberg: Springer, 2005: 102–109.
[11]	张强, 郝凯, 李海滨. 水下环境中基于曲线约束的SIFT特征匹配算法研究[J]. 光学学报, 2014, 34(2): 0215003-1−0215003-7. doi: 10.3788/AOS201434.0215003 ZHANG Q, HAO K, LI H B. Research on scale invariant feature transform feature matching based on underwater curve constraint[J]. Acta Optica Sinica, 2014, 34(2): 0215003-1−0215003-7 (in Chinese). doi: 10.3788/AOS201434.0215003
[12]	李佳宽, 孙春生, 胡艺铭, 等. 一种基于ORB特征的水下立体匹配方法[J]. 光电工程, 2019, 46(4): 180456. LI J K, SUN C S, HU Y M, et al. An underwater stereo matching method based on ORB features[J]. Opto-Electronic Engineering, 2019, 46(4): 180456 (in Chinese).
[13]	CHEN H H, WANG C C, SHIU D C, et al. A preliminary study on positioning of an underwater vehicle based on feature matching of seafloor images[C]//2018 OCEANS-MTS/IEEE Kobe Techno-Oceans (OTO). Kobe, Japan: IEEE, 2018: 1–6.
[14]	ROSTEN E, PORTER R, DRUMMOND T. Faster and better: a machine learning approach to corner detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010, 32(1): 105–119. doi: 10.1109/TPAMI.2008.275
[15]	张岩松. 基于深度约束的水下立体匹配研究[D]. 秦皇岛: 燕山大学, 2017. ZHANG Y S. Researching on underwater stereo matching based on depth constraint[D]. Qinhuangdao: Yanshan University, 2017 (in Chinese).
[16]	刘士荣, 吴楚, 张波涛, 等. 基于极线约束SIFT特征和粒子滤波的目标跟踪算法[J]. 上海交通大学学报, 2014, 48(7): 1026–1032,1038. LIU S R, WU C, ZHANG B T, et al. A tracking algorithm based on SIFT feature and particle filter with epipolar constraint[J]. Journal of Shanghai Jiaotong University, 2014, 48(7): 1026–1032,1038 (in Chinese).
[17]	李炼, 李维嘉, 吴耀中. 基于红色暗通道先验理论与CLAHE算法的水下图像增强算法[J]. 中国舰船研究, 2019, 14(增刊 1): 175–182. doi: 10.19693/j.issn.1673-3185.01508 LI L, LI W J, WU Y Z. An underwater image enhancement algorithm based on RDCP and CLAHE[J]. Chinese Journal of Ship Research, 2019, 14(Supp 1): 175–182 (in Chinese). doi: 10.19693/j.issn.1673-3185.01508
[18]	SKINNER K A, ZHANG J M, OLSON E A, et al. UWStereoNet: unsupervised learning for depth estimation and color correction of underwater stereo imagery[C]//Proceedings of 2019 International Conference on Robotics and Automation (ICRA). Montreal, QC, Canada: IEEE, 2019: 7947–7954.
[19]	SWIRSKI Y, SCHECHNER Y Y. 3Deflicker from motion[C]//Proceedings of IEEE International Conference on Computational Photography (ICCP). Cambridge, MA, USA: IEEE, 2013: 1–9.