A novel approach for fault detection in the aircraft body using image processing

This paper presents a novel method to inspect an aircraft structure and provide update regarding the maintenance. The purpose of this work is to automate the whole process of detection of faults in aircraft body using images and then to compare it with functioning images of the aircraft and concluding whether that section of the aircraft is faulty and needs maintenance. The images are taken by the robot moving on the ground. The idea of using ground moving robot is inspired from automated warehouses that uses automated robots to perform the tasks efficiently. Similarly, automated robot is prepared that will follow the prescribed path and take the images of the aircraft. The proposed idea will reshape the aircraft inspection by significantly reducing the periodic inspection time. So, the aircraft will be inspected periodically before and after the flight and supposedly its visits to hangar and maintenance pits will be reduced. This study sampled processing images of the outside of the aircraft, and the Convolutional Neural Network (CNN) approach uses features from the images to collect distinguishing features from a single patch created by the frame segmentation of a CNN kernel. Moreover, different filters are used to process the images using the toolbox for PYTHON image processing. At the initial run, it is observed that CNN falls for the overfitting of the faulty class. So, to overcome this problem image augmentation is applied and small dataset of 87 images is converted to the augmented dataset of 4000 images. After passing the data through several convolutional layers and multiple epochs execution the proposed model achieved the training accuracy of 98.28%.