General Damage Detections on Composite Panels Using Computer Vision Algorithms

This study aims to present a novel approach for composite panel inspection to identify Barely Visible Impact Damage (BVID) and Visible Impact Damage (VID), using computer vision algorithms. A new methodology for visual inspections was developed using computer vision algorithms is based on the YOLO - You Only Look Once architecture about the aircraft composite components, that use digital optics and deep learning techniques to identify and classify damages. Computer vision algorithms are used as instruments for automating the process of surface inspection and damage detection, increasing productivity and safety for maintenance operators (in the case of inspection of remote areas of aircraft). An overview of the problems, methods, and recent developments in deep learning algorithms used for general damage detection is provided. Data for Convolutional Neural Network (CNN) training were collected using a high-quality acquisition system. The database was populated by collecting images with damages from many composite panels, previously damaged by different impactors. Each defect has been photographed under different lighting conditions (bright or dark), resolution, and lighting angles to accurately simulate the environmental conditions of an aircraft maintenance hangar.