- This event has passed.
Thesis Defence: Registration and Fusion of Aircraft Lightning Damage NDE Data
August 10, 2023 at 11:00 am - 2:00 pm
Yanshuo Fan, supervised by Dr. Zheng Liu, will defend their thesis titled “Registration and Fusion of Aircraft Lightning Damage NDE Data” in partial fulfillment of the requirements for the degree of Master of Applied Science in Electrical Engineering.
An abstract for Yanshuo Fan’s thesis is included below.
Defences are open to all members of the campus community as well as the general public. Please email firstname.lastname@example.org to receive the Zoom link for this defence.
Lightning strikes are a critical concern for aircraft safety, necessitating the evaluation and maintenance of aircraft surface materials after each lightning event. Non-destructive evaluation/testing/inspection (NDE/NDT/NDI) methods play a vital role in assessing the structural condition of aircraft during maintenance procedures. However, each NDE technique has strengths and limitations when evaluating the stroked material. Therefore, the research in the thesis introduces novel image registration and fusion techniques that aim to overcome the limitations of individual NDE techniques and provide a comprehensive approach to aircraft material assessment.
First, a structurally enhanced NDE image registration model is proposed, which can automatically align NDE images with the ground truth images. With the structural enhancement filter, the registration accuracy improves significantly. Compared to existing advanced models, the proposed model achieves reliable and consistent alignment results on multi-modal NDE images without training.
Then, a two-stage image fusion-translation network (FTnet) is proposed. This network combines an image fusion module with an image translation module to enhance the quality of the fused images. The fusion module employs a novel fusion strategy that yields excellent results despite its simple structure. Moreover, the image translation module optimizes the resulting image representation without requiring additional training. The fused images demonstrate a remarkable enhancement of the image information. FTnet surpasses existing methods without relying on large datasets. The model signifies its substantial advantages in the field of image fusion.
The most remarkable advantage of the proposed models is their ability to provide reliable and efficient results in real industries with less reliance on large datasets. This makes these models more practical in the industry, where obtaining large amounts of data is challenging. Thus, the proposed models provide a promising way to improve NDE image results in industrial environments while minimizing data requirements.