지원사업
학술연구/단체지원/교육 등 연구자 활동을 지속하도록 DBpia가 지원하고 있어요.
커뮤니티
연구자들이 자신의 연구와 전문성을 널리 알리고, 새로운 협력의 기회를 만들 수 있는 네트워킹 공간이에요.
이용수0
1 Introduction 11.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Simultaneous Localization and Mapping . . . . . . . . . . . . . . . . 31.3 Visual Loop Closure Detection . . . . . . . . . . . . . . . . . . . . . 41.4 Problem Statement and Challenges . . . . . . . . . . . . . . . . . . . 51.5 Thesis Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 71.6 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 Literature Review 112.1 Features Based Loop Closure Detection Methods . . . . . . . . . . . 132.2 Deep Learning Based Loop Closure Detection Methods . . . . . . . . 162.3 Semantics Based Loop Closure Detection Methods . . . . . . . . . . 182.4 Open Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202.4.1 Viewpoint and Conditional Variations . . . . . . . . . . . . . 202.4.2 Dynamic Interference . . . . . . . . . . . . . . . . . . . . . . 212.4.3 Real-Time Performance . . . . . . . . . . . . . . . . . . . . . 222.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223 Binary Features Based Visual Loop Closure Detection 253.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.2 Features Based Visual Loop Closure Detection Techniques . . . . . . 293.3 Robust Feature Matching Method Based Visual Loop Closure Detection 313.3.1 Feature Extraction . . . . . . . . . . . . . . . . . . . . . . . . 333.3.1.1 Feature Detectors . . . . . . . . . . . . . . . . . . . 333.3.1.2 Feature Descriptors . . . . . . . . . . . . . . . . . . 333.3.2 BoW Vocabulary Generation . . . . . . . . . . . . . . . . . . 343.3.3 Vocabulary Based Image Matching . . . . . . . . . . . . . . . 353.3.3.1 Image to BoW Vector Conversion . . . . . . . . . . 353.3.3.2 BoW Vectors Matching: . . . . . . . . . . . . . . . . 353.3.4 Robust Feature Matching . . . . . . . . . . . . . . . . . . . . 373.3.4.1 Nearest Neighbor Symmetric Match . . . . . . . . . 373.3.4.2 Block Matching . . . . . . . . . . . . . . . . . . . . 383.3.5 Geometric Consistency . . . . . . . . . . . . . . . . . . . . . . 393.3.6 Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . 403.4 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . 403.4.1 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . 403.4.2 Evaluation Datasets . . . . . . . . . . . . . . . . . . . . . . . 413.4.3 Evaluation Parameters . . . . . . . . . . . . . . . . . . . . . . 433.4.3.1 Correct Matches (%) . . . . . . . . . . . . . . . . . 433.4.3.2 Precision-Recall Curve . . . . . . . . . . . . . . . . 433.4.3.3 Execution Time . . . . . . . . . . . . . . . . . . . . 433.5 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 443.5.1 No. of Extracted Features Per Image . . . . . . . . . . . . . . 443.5.2 Performance Evaluation of Feature Extraction Algorithms . . 443.5.3 Comparison of the Proposed Method with State-Of-The-ArtAlgorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473.5.4 Execution Time . . . . . . . . . . . . . . . . . . . . . . . . . . 503.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514 Semantic Global and Local Descriptor Based Visual Loop ClosureDetection 534.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534.2 Semantics Aided Visual Loop Closure Detection Method . . . . . . . 554.2.1 Visual Information Extraction . . . . . . . . . . . . . . . . . 574.2.2 Coarse Loop Closure Detection . . . . . . . . . . . . . . . . . 574.2.3 Fine Loop Closure Detection . . . . . . . . . . . . . . . . . . 594.2.3.1 Semantically Salient Local Descriptor Matching . . 594.2.3.2 Semantically Aggregated Global Descriptor Matching 604.2.3.3 Fusion Model . . . . . . . . . . . . . . . . . . . . . . 624.3 Experiments and Results . . . . . . . . . . . . . . . . . . . . . . . . . 624.3.1 Implementation Details . . . . . . . . . . . . . . . . . . . . . 624.3.2 Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 634.3.2.1 Oxford RobotCar Dataset . . . . . . . . . . . . . . . 634.3.2.2 Synthia Dataset . . . . . . . . . . . . . . . . . . . . 644.3.2.3 Mapillary Dataset . . . . . . . . . . . . . . . . . . . 644.3.3 Evaluation Parameters . . . . . . . . . . . . . . . . . . . . . . 644.3.4 Ablation Study . . . . . . . . . . . . . . . . . . . . . . . . . . 654.3.4.1 Top Candidates Selection . . . . . . . . . . . . . . . 654.3.4.2 Comparison with Baseline Algorithm . . . . . . . . 674.3.5 Comparison with State-Of-The-Art Methods . . . . . . . . . 674.3.6 Runtime Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 704.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 725 BoSVW-SGLD Based Visual Loop Closure Detection 755.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 755.1.1 Low Level Feature Extraction for Static Objects . . . . . . . 755.1.2 Semantics Aided Coarse Loop Candidates’ Selection . . . . . 765.1.3 Semantic Verification for Robust Feature Matching . . . . . . 765.1.4 Global Scene Understanding . . . . . . . . . . . . . . . . . . . 765.1.5 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 775.2 Visual and Semantic Information Extraction . . . . . . . . . . . . . . 795.2.1 Low Level Binary Features . . . . . . . . . . . . . . . . . . . 795.2.2 High Level Semantics . . . . . . . . . . . . . . . . . . . . . . 805.3 Hierarchical Bag-of-Semantic-Visual-Words Model . . . . . . . . . . 805.3.1 Robust Semantic-Visual Features Extraction . . . . . . . . . 825.3.2 A Semantic-Visual Words Vocabulary Tree . . . . . . . . . . 855.3.3 Coarse Loop Candidate Detection . . . . . . . . . . . . . . . 875.4 Semantic Global and Local Descriptor Similarity Fusion Model . . . 895.4.1 Semantic Global Descriptor Based Scene Matching . . . . . . 895.4.1.1 Semantic Global Descriptor Computation . . . . . . 905.4.1.2 Semantic Global Descriptor Matching . . . . . . . . 915.4.2 Spatially and Semantically Consistent Robust Local Descriptor Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . 915.4.2.1 Robust Feature Matching For Static Objects . . . . 925.4.2.2 Spatially Consistent Robust Feature Selection . . . 935.4.2.3 Semantic Inconsistency Filtration . . . . . . . . . . 945.4.2.4 Geometric Verification . . . . . . . . . . . . . . . . . 945.4.3 Fusion Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 955.5 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 955.5.1 Implementation Setup . . . . . . . . . . . . . . . . . . . . . . 955.5.2 Evaluation Datasets . . . . . . . . . . . . . . . . . . . . . . . 965.5.2.1 Oxford RobotCar Dataset . . . . . . . . . . . . . . . 965.5.2.2 Synthia Dataset . . . . . . . . . . . . . . . . . . . . 975.5.2.3 Mapillary Dataset . . . . . . . . . . . . . . . . . . . 975.5.3 Evaluation Parameters . . . . . . . . . . . . . . . . . . . . . . 975.5.3.1 True Positive Detection Rate / Hit rate . . . . . . . 985.5.3.2 Precision-Recall Curve . . . . . . . . . . . . . . . . 985.5.4 Area Under Precision-Recall Curve . . . . . . . . . . . . . . . 985.5.5 Ablation Study . . . . . . . . . . . . . . . . . . . . . . . . . . 995.5.6 Comparison With Baseline Algorithm . . . . . . . . . . . . . 995.5.7 Comparison With State-Of-The-Art Algorithms . . . . . . . . 1015.5.8 Runtime Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 1045.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1096 Conclusions, Applications and Future Research Directions 1116.1 Contributions Summary . . . . . . . . . . . . . . . . . . . . . . . . . 1126.2 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1146.2.1 Autonomous Mobile Robot Navigation . . . . . . . . . . . . . 1146.2.2 Delivery Service Robots . . . . . . . . . . . . . . . . . . . . . 1156.2.3 Multi Agent SLAM . . . . . . . . . . . . . . . . . . . . . . . . 1156.3 Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1166.3.1 Detection Performance Improvement . . . . . . . . . . . . . . 1166.3.2 Computational Efficiency . . . . . . . . . . . . . . . . . . . . 117List of Author’s Publications 119Bibliography 123
0