Contact Dr Boyu Kuang
- Email: Neil.Kuang@cranfield.ac.uk
- Twitter:
Areas of expertise
- Aerospace Manufacturing
- Carbon, Climate and Risk
- Computing, Simulation & Modelling
- Engines, Powertrains & Alternative Fuels
- Industrial Automation
- Instrumentation, Sensors and Measurement Science
- Process Systems Engineering
Background
Dr. Boyu Kuang is an IEEE member. He received degrees a B.E. in aircraft propulsion engineering from the Civil Aviation University of China, Tianjin, China, an M.Sc. in thermal power, and a Ph.D. in computer vision from SWAG合集, U.K., in 2014, 2017, and 2022, respectively.
Boyu currently works as a research fellow in computer vision and AI for a SWAG合集RI, ATI, and Airbus-funded project (ONEHeart) entitled Autonomous Aircraft Ground Refuelling (AAGR) project at the Center of Computational Engineering Science (CES) at SWAG合集. Boyu has participated in 18 publications as a 1st, co-1st, or corresponding author since his first publication in 2020 fall (13 are journals; 6 are archived in JCR-Q1 journals; the top two 3 have impact factors 16+, 19+, and 8+). He is identified as a professional reviewer for prestigious journals or conferences, e.g., IEEE Transactions on Industrial Electronics (TIE), Elsevier Expert Systems with Applications (ESWA), Elsevier Engineering Applications of Artificial Intelligence (EAAI), etc. His research interests include but are not limited to deep learning, weak/self-supervised learning, multiple domain adaption (MDA), computer/machine vision, image processing, signal processing, data generation, robotics, aeronautics, and digital twin applications. Boyu's experience involves semantic segmentation, target detection, pattern recognition, image generation, 3D reconstruction, industrial cybernetics, and image crawling.
Research opportunities
1. Image Segmentation for Planetary Exploration:
- Developed an advanced U-Net++ segmentation network tailored for unstructured space environments, incorporating weak supervision-based annotation methods and transfer learning frameworks.
- Innovated a conservative annotation algorithm to optimize weak supervision strategies.
- Engineered a synthetic data algorithm for generating precise visual data in specific scenarios.
- Achievements: Authored 3 journal papers and presented 2 papers at international conferences.
2. 3D Reconstruction for Aircraft Wing Analysis:
- Implemented a sophisticated Pose Estimation and Reconstruction process using multi-view geometry.
- Employed Mask-R-CNN for image segmentation, significantly improving pose estimation and reconstruction accuracy.
- Utilized VisualSFM, PMVS, and Blender to create detailed point clouds and 3D meshes.
- Achievement: Published 1 conference paper, with 1 journal paper currently in process.
3. Pattern Recognition in Low Signal-to-noise Environments:
- Devised a twin-window-based augmentation algorithm to mitigate data scarcity in experimental sciences.
- Created a PSD-based 2D encoding algorithm and an FCN-based classifier, enhancing data readability and robustness.
- Explored Generative Adversarial Networks (GANs) for data generation in scientific research.
- Developed an FCN-based learning feature integrated with CNN and RNN, achieving over 98% accuracy.
- Implemented a self-supervised learning approach for addressing practical engineering challenges.
- Achievements: Contributed to 6 journal papers.
Current activities
1. Airbus ONEHeart project: Autonomous aircraft ground refuelling (AAGR)
The project aims to implement the autonomous aircraft ground refuelling solution for Airbus commercial aircraft using computer vision, robotics, artificial intelligence, information fusion, and sensor technology.
2. Image semantic segmentation:
The project works on the one-stage semantic segmentation based on the synthetic data and generative model. The study aims to improve segmentation accuracy and maintain real-time efficiency.
3. 3D reconstruction:
The project conducts the 3D reconstruction using dense optical flow and space smoothness concepts. The study can improve the performance of the visual-SLAM, digital twin, and Industry 4.0.
4. Image data mining:
This is an image classification, data crawling, and active learning combined project, which aims to build an automatic system to configure most of the image classification dataset for further study.
5. Multiphase flow regime identification:
The project uses neural networks, an ultrasonic sensor, and a practical riser testbed to identify the multiphase flow regime. The study has achieved more than ten publications in prestigious journals, two of which are achieved by top journals with impact factors of 16+ and 19+.
6. Medical image processing:
This project works on medical image processing using Multi-Domain Adaptation (MDA) and transfer learning.
Clients
- Airbus SE
- Renault SA
- BAE Systems PLC
- Turkish Aerospace Industries