Nondestructive testing application of Cessna 172R aircraft skin coating based on spectral indices

Damage to the skin coating of the Cessna 172R aircraft is an unavoidable and significant issue due to long flight training. Traditional detection methods are easy to missed and false detections, and hyperspectral technology can significantly improve. The spectral curves of the damaged and undamaged skin coating pixels in the near-infrared band (900-1700 nm) of the Cessna 172R aircraft skin samples were used to establish three spectral indices: ASCI-I, ASCI-II, and ASCI-III. Then the decision tree is used to perform recognition experiments on the two types of skin samples. The experimental results show that the decision tree model based on ASCI-I has the best recognition performance. Its global image recognition results are closely related to the spatial distribution of the actual targets. The Producer accuracy, User accuracy, and overall classification accuracy are all over 90%, and the Kappa coefficient is greater than 0.88.