Design Of Ai-Driven Computer Vision Software For Aeronautical Testing

Abstract

In the aeronautical industry, the safety, functionality, and reliability of aircraft systems are fundamental. Traditional testing processes are often time-consuming and resource-intensive. This paper presents the design and implementation of AI-driven computer vision software aimed at optimizing these testing processes through automation. By leveraging advanced Convolutional Neural Networks (CNNs) and Optical Character Recognition (OCR) algorithms, the software can perform real-time detection and post-flight video analysis. Specifically, a YOLOv8 model has been fine-tuned for object detection tasks within various testing phases, including unit tests, system integration tests, and flight tests. The integration of PaddleOCR enhances the software functionalities adding text recognition capabilities. This approach not only speeds up the testing process but also significantly reduces human effort to manually collect data, leading to improved efficiency and human errors. The paper further explores the implementation details, challenges, and performance metrics of the applied software, demonstrating their potential to improve aeronautical testing procedures.