Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor | Department of Aeronautical and Aviation Engineering | en_US |
| dc.contributor.advisor | Wen, Chih-yung (AAE) | en_US |
| dc.creator | Yang, Wenyu | - |
| dc.identifier.uri | https://theses.lib.polyu.edu.hk/handle/200/14537 | - |
| dc.language | English | en_US |
| dc.publisher | Hong Kong Polytechnic University | en_US |
| dc.rights | All rights reserved | en_US |
| dc.title | Multi-sensor fusion based intelligent perception using UAVs for autonomous infrastructure and agriculture inspection | en_US |
| dcterms.abstract | With the rapid advancement of Artificial Intelligence (AI) and autonomous technologies, intelligent robotics is spearheading a profound technological revolution, comprehensively reshaping our way of life. Significant benefits from technological innovation have already been realized across multiple domains, including autonomous driving, intelligent vehicles for specialized scenarios (factories, mines, farms), and intelligent aerial robots. | en_US |
| dcterms.abstract | Nevertheless, numerous open challenges remain to be addressed in the coming decades. Examples include the ongoing paradigm shift in autonomous driving—from modular design to end-to-end approaches—and the realization of autonomous robot operation across all terrains and throughout the full lifecycle. Amidst these challenges, burgeoning fields such as Embodied AI, humanoid robotics, and aerial manipulation are rapidly emerging. | en_US |
| dcterms.abstract | Despite the existence of diverse research directions in the field, intelligent perception stands as the fundamental cornerstone of intelligent robot systems. Only with accurate and robust intelligent perception capabilities can an intelligent robot system establish a reliable foundation, thereby enabling the seamless execution of downstream tasks such as motion planning and precise control. In this context, our research focuses on intelligent perception tasks driven by multi-sensor fusion, which achieves centimeter-level translation accuracy and sub-degree level rotation accuracy. We further extend our work to the development of a multi-sensor-based Simultaneous Localization and Mapping (SLAM) system, as well as the advancement of LiDAR Bundle Adjustment and hierarchical 3D scene graph techniques. This research work encompasses four primary components: | en_US |
| dcterms.abstract | • Multi-sensor Calibration Algorithms: Development of methods for accurately calibrating diverse sensors. | en_US |
| dcterms.abstract | • SLAM Front-end: Implementation of the multi-sensor fusion SLAM system for online pose estimation. | en_US |
| dcterms.abstract | • LiDAR Bundle Adjustment for Accurate Geometric Mapping: Uncertainty-based optimization techniques are applied to LiDAR data to improve the geometric accuracy of the constructed maps. | en_US |
| dcterms.abstract | • Hierarchical 3D Scene Graphs for Metric-Semantic Mapping: Construction of hierarchical representations integrating geometric and semantic information within the environment. | en_US |
| dcterms.abstract | We employ Unmanned Aerial Vehicle (UAV) as the mobile platform, validating the proposed system's efficacy through two case studies while adapting it to the specific characteristics of distinct application scenarios and tasks. The first application scenario is intelligent vertical farming, and the second is bridge inspection. | en_US |
| dcterms.abstract | In the vertical farming application, we monitor the growth status of crops (strawberries) and perform 3D mapping and counting of strawberry blossoms and fruits. Building upon traditional geometric point cloud maps, we introduce three-dimensional Scene Graph (3DSG) to manage semantic information across different levels of abstraction. By co-optimizing parent and child nodes within the scene graph, our method achieves significantly enhanced fruit counting accuracy compared to conventional approaches. | en_US |
| dcterms.abstract | In the bridge management application, we continue utilizing scene graphs to organize semantic cues critical for bridge inspection and management tasks. Furthermore, tailored to the structural characteristics of bridges, we incorporate corner points, edges, and planar features into the LiDAR Bundle Adjustment (BA) process, which substantially improves the precision of LiDAR mapping. | en_US |
| dcterms.abstract | In conclusion, this study comprehensively demonstrates the considerable potential of UAV-based intelligent inspection systems in agricultural and industrial applications, encompassing sensor fusion, SLAM, semantic detection, 3D object estimation, and high-level organizational structures for management purposes. By integrating AI components, the system significantly enhances the effectiveness of UAVs in infrastructure inspection, providing valuable insights for practitioners in agricultural monitoring and structural health monitoring. Extensive simulations and real-world flight tests validate the system's efficacy, bringing the vision of intelligent inspection systems closer to reality and laying a solid foundation for downstream tasks. | en_US |
| dcterms.extent | xviii, 131 pages : color illustrations | en_US |
| dcterms.isPartOf | PolyU Electronic Theses | en_US |
| dcterms.issued | 2026 | en_US |
| dcterms.educationalLevel | Ph.D. | en_US |
| dcterms.educationalLevel | All Doctorate | en_US |
| dcterms.LCSH | Drone aircraft | en_US |
| dcterms.LCSH | Multisensor data fusion | en_US |
| dcterms.LCSH | Structural health monitoring | en_US |
| dcterms.LCSH | Precision farming | en_US |
| dcterms.LCSH | Hong Kong Polytechnic University -- Dissertations | en_US |
| dcterms.accessRights | open access | en_US |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/14537