Next >

Visual Perception and Control of Underwater Robots

Research Background Review of Underwater Visual Restoration Formation of Underwater Image Visual Restoration Based on Image Formation Model Visual Restoration Based on Information Fusion Review of Deep-Learning-Based Object Detection Two-Stage Detector RCNN Fast RCNN Faster RCNN RFCN Single-Stage Detector YOLO SSD RetinaNet RefineDet Temporal Object Detection Post-Processing Cascade of Detection and Tracking Feature Fusion Based on Motion Estimation Feature Propagation Based on RNN Temporally Sustained Proposal Batch-Processing Benchmarks of Object Detection PASCAL VOC MS COCO ImageNet VID Evaluation Metrics REVIEW OF UNDERWATER STEREO MEASUREMENT Overview of the Subsequence Chapters References Adaptive Real-Time Underwater Visual Restoration with Adversarial Critical Learning Introduction Review of Visual Restoration and Image-to-Image Translation Traditional Underwater Image Restoration Methods Image-to-Image Translation GAN-Based Restoration with Adversarial Critical Learning Filtering-Based Restoration Scheme Architecture of the GAN-Based Restoration Scheme Objective for GAN-RS Adversarial Loss DCP Loss Underwater Index Loss Full Loss Experiments and Discussion Details of ACL Basic Settings Multistage Loss Strategy Compared Methods Runtime Performance Running Environment Time Efficiency Restoration Results Visualization of Underwater Index Comparison on Restoration Quality Feature-Extraction Tests Visualization of Discriminator Discussion Concluding Remarks References A NSGA-II-Based Calibration for Underwater Binocular Vision Measurement Introduction Related Work Refractive Camera Model Akin Triangulation and Refractive Constraint Akin Triangulation Refractive Surface Constraint Calibration Algorithm A Novel Usage of Checkerboard Analysis of the Binocular Housing Parameters NSGA-II Algorithm Process of the Calibration Algorithm Experiments and Results Experimental Setup Results of Calibration Experiments on Position Measurement Experiments on Position Measurement Discussion Conclusion and Future Work References Joint Anchor-Feature Refinement for Real-Time Accurate Object Detection in Images and Videos Introduction Review of Deep Learning-Based Object Detection CNN-Based Static Object Detection Temporal Object Detection Sampling for Object Detection Dual Refinement Network Overall Architecture Anchor-Offset Detection From SSD to RefineDet, then to DRNet Anchor Refinement Deformable Detection Head Feature Location Refinement Multi-deformable Head Training and Inference Temporal Dual Refinement Networks Architecture Training Inference Experiments and Discussion Ablation Studies of DRNet320-VGG16 on VOC 2007 Anchor-Offset Detection Multi-deformable Head Toward More Effective Training Results on VOC 2007 Results on VOC 2012 Results on COCO Results on ImageNet VID Accuracy vs. Speed Trade-off Comparison with Other Architectures Discussion Concluding Remarks References Rethinking Temporal Object Detection from Robotic Perspectives Introduction Review of Temporal Detection and Tracking On VID Temporal Performance Non-reference Assessments Recall Continuity Localization Stability Online Tracklet Refinement SOT-by-Detection Small-Overlap Suppression SOT-by-Detection Framework Experiments and Discussion Analysis on VID Continuity/Stability Tracklet Visualization Numerical Evaluation SOT-by-Detection Speed Comparison of NMS and SOS-NMS SOT-by-Detection vs. Siamese SOT Discussion Concluding Remarks References Reveal of Domain Effect: How Visual Restoration Contributes to Object Detection in Aquatic Scenes Introduction Review of Underwater Visual Restoration and Domain-Adaptive Object Detection Underwater Visual Restoration Domain-Adaptive Object Detection Preliminary Preliminary of Data Domain Based on Visual Restoration Domain Generation Domain Analysis Preliminary of Detector Joint Analysis on Visual Restoration and Object Detection Within-Domain Performance Numerical Analysis Visualization of Convolutional Representation Precision-Recall Analysis Cross-Domain Performance Cross-Domain Evaluation Cross-Domain Training Domain Effect on Real-World Object Detection Online Object Detection in Aquatic Scenes Online Domain Analysis Discussion Underwater Vision System and Marine Test System Design Underwater Object Counting Underwater Object Grasping Concluding Remarks References IWSCR: An Intelligent Water Surface Cleaner Robot for Collecting Floating Garbage Introduction Prototype Design of IWSCR Configuration of IWSCR Framework of Control System ACCURATE AND REAL-TIME GARBAGE DETECTION Sliding Mode Controller for Vision-Based Steering Dynamic Model of Underwater Vehicle Formulation of the Vision-Based Steering Design and Stability Analysis of Sliding Mode Controller Dynamic Grasping Strategy for Floating Bottles Kinematics and Inverse Kinematics of Manipulator Description of the Feasible Grasping Strategy Experiments and Discussion Experimental Results of Garbage Detection Experimental Results of SMC for Vision-Based Steering and Achievement of TTs Discussion CONCLUSION AND FUTURE WORK References Underwater Target Tracking Control of an Untethered Robotic Fish with a Camera Stabilizer Introduction System Design of the Robotic Fish with a Camera Stabilizer Mechatronic Design CPG-Based Motion Control Active Vision Tracking System RL-Based Target Tracking Control Tracking Control Design Performance Analysis of DDPG-Based Control System Experiments and Results Static and Dynamic Tracking Experiments Discussion Conclusions and Future Work References

Next >