AI-driven UAV with image processing algorithm for automatic visual inspection of aircraft external surface
Abstract This paper presents a novel AI-driven drone for automatic visual inspection based defects detection in the aircraft external surfaces. The defects on the aircraft surface are usually mixed with noise that are coming from unexpected sources such as aircraft’s background, the appearance of ri...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-06-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-025-02902-2 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract This paper presents a novel AI-driven drone for automatic visual inspection based defects detection in the aircraft external surfaces. The defects on the aircraft surface are usually mixed with noise that are coming from unexpected sources such as aircraft’s background, the appearance of rivet on the aircraft’s surface and the surrounding environment like non-homogeneity of light intensity, shadow and weather changing, leading to difficulty in distinguishing between the defects and noise by merely applying an image processing algorithm. Thus, an AI algorithm with capability to deal with noise has been introduced to properly classify the defects. The proposed AI algorithm consists of two subsequent stages with a novel algorithm, called optimized laser simulator logic that is capable to accommodate the noise by applying a high degree of overlapping between the linguistic variables and make the right decision on the defects. The results show that the image processing techniques are effective in extracting features of possible defects such as cracks, dents, and scratches in samples image of aircraft surfaces. Meanwhile, the two stages of AI-algorithm demonstrate a good capability on classifying the extracted features by image processing into possible defect or noises which yields to accuracy rates of 86.67%, 66.67%, 80.0%, and 76.67% for cracks, dents, scratches, and rust, respectively. The proposed AI algorithm has been compared with Yolo 11 trained on ROBOFLOW dataset, which shows that the proposed algorithm outperforms Yolo 11 in terms of precision, recall, F-score and accuracy metrics. The proposed system will shorten the waiting time to accomplish the pre-flight checks in airports. |
|---|---|
| ISSN: | 2045-2322 |