Deep Learning Based Object Recognition Approach for Motorbike Dashboard Software Validation

Abstract

With the increasing complexity of motorbike dashboard software, ensuring its reliability and functionality has become a significant challenge in the automotive industry. Traditional manual testing methods are labor intensive, time consuming and prone to human error. This study explores the application of deep learning based object detection techniques to automate dashboard software validation. To implement the proposed testing environment, a camera system is integrated with a hardware in the loop setup. The system captures and processes dashboard images using a deep learning model, which detects key elements and compares them to reference values for validation. This approach significantly enhances testing accuracy and repeatability. The study begins by reviewing traditional object detection methods and their limitations. It then examines modern deep learning based techniques, with a focus on one-stage and two-stage detectors. One-stage detectors are found to offer a balance between speed and accuracy which makes them suitable for dashboard validation. Additionally, the study highlights the importance of camera system selection, it was found that high-resolution and high dynamic range sensors improve detection performance. Other factors such as dataset preparation, environment control and software architecture are also discussed in detail. The proposed methodology is structured around three milestones. It covers the implementation of the test setup, the training of the detection model and the final system integration. A modular and scalable standalone application was developed to handle image capturing, detection and result comparison. Additionally, it was successfully integrated with a hardware in the loop setup for comprehensive testing. The findings of this study demonstrate that deep learning based object detection, combined with optimized camera system and hardware in the loop setup, presents a reliable and scalable solution for dashboard software validation. Future research may focus on expanding the dataset, adding new validation layers and enabling real time processing.

Date of Award	2025
Original language	English
Supervisor	Thomas Schlechter (Supervisor)