Muutke küpsiste eelistusi

Visual Perception and Control of Underwater Robots [Kõva köide]

, ,
  • Formaat: Hardback, 236 pages, kõrgus x laius: 229x152 mm, kaal: 471 g, 21 Tables, black and white; 56 Line drawings, black and white; 49 Halftones, black and white; 105 Illustrations, black and white
  • Ilmumisaeg: 01-Mar-2021
  • Kirjastus: CRC Press
  • ISBN-10: 0367695782
  • ISBN-13: 9780367695781
Teised raamatud teemal:
Visual Perception and Control of Underwater RobotsSuurem pilt
  • Formaat: Hardback, 236 pages, kõrgus x laius: 229x152 mm, kaal: 471 g, 21 Tables, black and white; 56 Line drawings, black and white; 49 Halftones, black and white; 105 Illustrations, black and white
  • Ilmumisaeg: 01-Mar-2021
  • Kirjastus: CRC Press
  • ISBN-10: 0367695782
  • ISBN-13: 9780367695781
Teised raamatud teemal:
Püsilink: https://www.kriso.ee/db/9780367695781.html
Märksõnad:
Visual Perception and Control of Underwater Robots covers theories and applications from aquatic visual perception and underwater robotics. Within the framework of visual perception for underwater operations, image restoration, binocular measurement, and object detection are addressed. More specifically, the book includes adversarial critic learning for visual restoration, NSGA-II-based calibration for binocular measurement, prior knowledge refinement for object detection, analysis of temporal detection performance, as well as the effect of the aquatic data domain on object detection.

With the aid of visual perception technologies, two up-to-date underwater robot systems are demonstrated. The first system focuses on underwater robotic operation for the task of object collection in the sea. The second is an untethered biomimetic robotic fish with a camera stabilizer, its control methods based on visual tracking.

The authors provide a self-contained and comprehensive guide to understand underwater visual perception and control. Bridging the gap between theory and practice in underwater vision, the book features implementable algorithms, numerical examples, and tests, where codes are publicly available. Additionally, the mainstream technologies covered in the book include deep learning, adversarial learning, evolutionary computation, robust control, and underwater bionics. Researchers, senior undergraduate and graduate students, and engineers dealing with underwater visual perception and control will benefit from this work.
Chapter 1 Introduction
1(32)
1.1 Research Background
1(3)
1.2 Review Of Underwater Visual Restoration
4(4)
1.2.1 Formation of Underwater Image
4(1)
1.2.2 Visual Restoration Based on Image Formation Model
5(1)
1.2.3 Visual Restoration Based on Information Fusion
6(2)
1.3 Review Of Deep-Learning-Based Object Detection
8(16)
1.3.1 Two-Stage Detector
8(1)
1.3.1.1 RCNN
8(2)
1.3.1.2 Fast RCNN
10(1)
1.3.1.3 Faster RCNN
10(1)
1.3.1.4 RFCN
10(1)
1.3.2 Single-Stage Detector
11(1)
1.3.2.1 YOLO
11(1)
1.3.2.2 SSD
12(3)
1.3.2.3 RetinaNet
15(1)
1.3.2.4 RefineDet
16(1)
1.3.3 Temporal Object Detection
16(1)
1.3.3.1 Post-Processing
17(1)
1.3.3.2 Cascade of Detection and Tracking
17(2)
1.3.3.3 Feature Fusion Based on Motion Estimation
19(1)
1.3.3.4 Feature Propagation Based on RNN
19(1)
1.3.3.5 Temporally Sustained Proposal
20(1)
1.3.3.6 Batch-Processing
21(1)
1.3.4 Benchmarks of Object Detection
22(1)
1.3.4.1 PASCAL VOC
22(1)
1.3.4.2 MS COCO
22(1)
1.3.4.3 ImageNet VID
23(1)
1.3.4.4 Evaluation Metrics
23(1)
1.4 Review Of Underwater Stereo Measurement
24(4)
1.5 Overview Of The Subsequence
Chapters
28(5)
References
28(5)
Chapter 2 Adaptive Real-Time Underwater Visual Restoration With Adversarial Critical Learning
33(24)
2.1 Introduction
33(3)
2.2 Review Of Visual Restoration And Image-To-Image Translation
36(1)
2.2.1 Traditional Underwater Image Restoration Methods
36(1)
2.2.2 Image-To-Image Translation
37(1)
2.3 Gan-Based Restoration With Adversarial Critical Learning
37(7)
2.3.1 Filtering-Based Restoration Scheme
38(1)
2.3.2 Architecture of the GAN-Based Restoration Scheme
39(2)
2.3.3 Objective for GAN-RS
41(1)
2.3.3.1 Adversarial Loss
41(1)
2.3.3.2 DCPLoss
41(1)
2.3.3.3 Underwater Index Loss
42(2)
2.3.3.4 Full Loss
44(1)
2.4 Experiments And Discussion
44(10)
2.4.1 Details of ACL
44(1)
2.4.1.1 Basic Settings
44(1)
2.4.1.2 Multistage Loss Strategy
44(1)
2.4.2 Compared Methods
45(1)
2.4.3 Runtime Performance
45(1)
2.4.3.1 Running Environment
45(1)
2.4.3.2 Time Efficiency
45(2)
2.4.4 Restoration Results
47(1)
2.4.4.1 Visualization of Underwater Index
47(1)
2.4.4.2 Comparison on Restoration Quality
47(2)
2.4.4.3 Feature-Extraction Tests
49(2)
2.4.5 Visualization of Discriminator
51(1)
2.4.6 Discussion
52(2)
2.5 Concluding Remarks
54(3)
References
54(3)
Chapter 3 A Nsga-II-Based Calibration For Underwater Binocular Vision Measurement
57(32)
3.1 Introduction
57(2)
3.2 Related Work
59(3)
3.3 Refractive Camera Model
62(3)
3.4 Akin Triangulation And Refractive Constraint
65(4)
3.4.1 Akin Triangulation
65(4)
3.4.2 Refractive Surface Constraint
69(1)
3.5 Calibration Algorithm
69(6)
3.5.1 A Novel Usage Of Checkerboard
70(2)
3.5.2 Analysis Of The Binocular Housing Parameters
72(1)
3.5.3 Nsga-11 Algorithm
72(3)
3.5.4 Process Of The Calibration Algorithm
75(1)
3.6 Experiments And Results
75(9)
3.6.1 Experimental Setup
75(1)
3.6.2 Results Of Calibration
76(3)
3.6.3 Experiments on Position Measurement
79(2)
3.6.4 Experiments on Position Measurement
81(1)
3.6.5 Discussion
81(3)
3.7 Conclusion And Future Work
84(5)
References
84(5)
Chapter 4 Joint Anchor-Feature Refinement For Real-Time Accurate Object Detection In Images And Videos
89(36)
4.1 Introduction
89(4)
4.2 Review Of Deep Learning-Based Object Detection
93(1)
4.2.1 Cnn-Based Static Object Detection
93(1)
4.2.2 Temporal Object Detection
93(1)
4.2.3 Sampling For Object Detection
94(1)
4.3 Dual Refinement Network
94(7)
4.3.1 Overall Architecture
95(1)
4.3.2 Anchor-Offset Detection
95(1)
4.3.2.1 From SSD to RefineDet, then to DRNet
95(3)
4.3.2.2 Anchor Refinement
98(1)
4.3.2.3 Deformable Detection Head
98(1)
4.3.2.4 Feature Location Refinement
98(1)
4.3.3 Multi-deformable Head
99(1)
4.3.4 Training and Inference
100(1)
4.4 Temporal Dual Refinement Networks
101(3)
4.4.1 Architecture
101(2)
4.4.2 Training
103(1)
4.4.3 Inference
103(1)
4.5 Experiments And Discussion
104(15)
4.5.1 Ablation Studies of DRNet320-VGG16 on VOC 2007
105(1)
4.5.1.1 Anchor-Offset Detection
105(2)
4.5.1.2 Multi-deformable Head
107(1)
4.5.1.3 Toward More Effective Training
108(1)
4.5.2 Results on VOC 2007
108(1)
4.5.3 Results on VOC 2012
108(2)
4.5.4 Results on COCO
110(4)
4.5.5 Results on ImageNet VID
114(1)
4.5.5.1 Accuracy vs. Speed Trade-off
114(2)
4.5.5.2 Comparison with Other Architectures
116(2)
4.5.6 Discussion
118(1)
4.5.6.1 Key Frame Scheduling
118(1)
4.5.6.2 Further Enhancement of Refinement Networks
118(1)
4.5.6.3 Refinement Networks for Real-World Object Detection
119(1)
4.6 Concluding Remarks
119(6)
References
120(5)
Chapter 5 Rethinking Temporal Object Detection from Robotic Perspectives
125(22)
5.1 Introduction
125(4)
5.2 Review Of Temporal Detection And Tracking
129(1)
5.2.1 Temporal Object Detection
129(1)
5.2.2 Tracking Metrics
129(1)
5.2.3 Tracking-by-Detection (i.e., MOT)
130(1)
5.2.4 Detection-SOT Cascade
130(1)
5.3 ON VID TEMPORAL PERFORMANCE
130(4)
5.3.1 Non-reference Assessments
130(1)
5.3.1.1 Recall Continuity
131(1)
5.3.1.2 Localization Stability
132(1)
5.3.2 Online Tracklet Refinement
133(1)
5.3.2.1 Short Tracklet Suppression
133(1)
5.3.2.2 Fragment Filling
133(1)
5.3.2.3 Temporal Location Fusion
134(1)
5.4 Sot-By-Detection
134(2)
5.4.1 Small-Overlap Suppression
134(1)
5.4.2 SOT-by-Detection Framework
135(1)
5.5 Experiments And Discussion
136(7)
5.5.1 Analysis on VID Continuity/Stability
137(1)
5.5.1.1 Tracklet Visualization
137(1)
5.5.1.2 Numerical Evaluation
137(4)
5.5.2 SOT-by-Detection
141(1)
5.5.2.1 Speed Comparison ofNMS and SOS-NMS
141(1)
5.5.2.2 SOT-by-Detection vs. Siamese SOT
141(1)
5.5.3 Discussion
142(1)
5.5.3.1 Detector-Based Improvement
142(1)
5.5.3.2 Limitation of SOT-by-Detection
142(1)
5.6 Concluding Remarks
143(4)
References
143(4)
Chapter 6 Reveal of Domain Effect: How Visual Restoration Contributes to Object Detection in Aquatic Scenes
147(24)
6.1 Introduction
147(3)
6.2 Review Of Underwater Visual Restoration And Domain-Adaptive Object Detection
150(1)
6.2.1 Underwater Visual Restoration
150(1)
6.2.2 Domain-Adaptive Object Detection
150(1)
6.3 Preliminary
151(3)
6.3.1 Preliminary of Data Domain Based on Visual Restoration
151(1)
6.3.1.1 Domain Generation
151(1)
6.3.1.2 Domain Analysis
152(1)
6.3.2 Preliminary of Detector
152(2)
6.4 Joint Analysis On Visual Restoration And Object Detection
154(9)
6.4.1 Within-Domain Performance
154(1)
6.4.1.1 Numerical Analysis
155(1)
6.4.1.2 Visualization of Convolutional Representation
155(1)
6.4.1.3 Precision-Recall Analysis
155(4)
6.4.2 Cross-Domain Performance
159(1)
6.4.2.1 Cross-Domain Evaluation
159(1)
6.4.2.2 Cross-Domain Training
160(1)
6.4.3 Domain Effect on Real-World Object Detection
161(1)
6.4.3.1 Online Object Detection in Aquatic Scenes
161(1)
6.4.3.2 Online Domain Analysis
162(1)
6.4.4 Discussion
163(1)
6.4.4.1 Recall Efficiency
163(1)
6.4.4.2 CNN's Domain Selectivity
163(1)
6.5 Underwater Vision System And Marine Test
163(3)
6.5.1 System Design
163(1)
6.5.2 Underwater Object Counting
164(1)
6.5.3 Underwater Object Grasping
165(1)
6.6 Concluding Remarks
166(5)
References
167(4)
Chapter 7 IWSCR: An Intelligent Water Surface Cleaner Robot for Collecting Floating Garbage
171(28)
7.1 Introduction
171(3)
7.2 Prototype Design Of Iwscr
174(2)
7.2.1 Configuration of IWSCR
174(1)
7.2.2 Framework of Control System
175(1)
7.3 Accurate And Real-Time Garbage Detection
176(1)
7.4 Sliding Mode Controller For Vision-Based Steering
177(8)
7.4.1 Dynamic Model of Underwater Vehicle
177(4)
7.4.2 Formulation of the Vision-Based Steering
181(2)
7.4.3 Design and Stability Analysis of Sliding Mode Controller
183(2)
7.5 Dynamic Grasping Strategy For Floating Bottles
185(3)
7.5.1 Kinematics and Inverse Kinematics of Manipulator
185(1)
7.5.2 Description of the Feasible Grasping Strategy
186(2)
7.6 Experiments And Discussion
188(7)
7.6.1 Experimental Results of Garbage Detection
188(1)
7.6.2 Experimental Results of SMC for Vision-Based Steering and Achievement of TTs
189(4)
7.6.3 Discussion
193(2)
7.7 Conclusion And Future Work
195(4)
References
195(4)
Chapter 8 Underwater Target Tracking Control Of An Untethered Robotic Fish With A Camera Stabilizer
199(34)
8.1 Introduction
199(3)
8.2 System Design Of The Robotic Fish With A Camera Stabilizer
202(4)
8.2.1 Mechatronic Design
202(2)
8.2.2 Cpg-Based Motion Control
204(2)
8.3 Active Vision Tracking System
206(5)
8.4 Rl-Based Target Tracking Control
211(10)
8.4.1 Tracking Control Design
211(5)
8.4.2 Performance Analysis Of Ddpg-Based Control System
216(5)
8.5 Experiments And Results
221(8)
8.5.1 Static And Dynamic Tracking Experiments
221(6)
8.5.2 Discussion
227(2)
8.6 Conclusions And Future Work
229(4)
References
229(4)
Chapter 9 Summary and Outlook
233
Junzhi Yu is a professor of Peking University, whose research interests incude biomimetic robots, intelligent control, and intelligent mechatonic systems. In these areas, he has (co-)authored 3 monographs, and published over 100 SCI papers in the prestigious robotics and automation related journals.

Xingyu Chen, PhD in University of Chinese Academy of Sciences.

Shihan Kong, PhD student in University of Chinese Academy of Sciences.