Advanced Driver Assistance Systems (ADAS) have become essential for improving driving safety and the overall driving experience in recent years. However, varying illumination conditions often lead to unstable image quality for practical driving situations. This study uses gamma correction and scene classification to explore methods for enhancing image stability under changing lighting conditions. Additionally, we refine the classification module for five specific brightness scenes and incorporate image enhancement modules tailored to address a wide range of illumination conditions, including low-light image enhancement and exposure correction. By dynamically adjusting the regression and leveraging advanced scene classifiers, we can accurately enhance local and global image details to adapt to varying lighting conditions. Our approach improves driver visibility, reliability, and safety in ADAS and lays the groundwork for developing high-level autonomous driving systems, encouraging further research and innovation in the field.

The challenges posed by dynamic lighting conditions and unpredictable scenes necessitate robust image enhancement techniques. Grounded in the theoretical framework of the atmospheric scattering model, our study introduces a novel image enhancement approach that leverages deep learning and is reinforced by the integration of gamma correction. The obtained numerical results demonstrate a significant improvement over current state-of-the-art techniques, effectively enhancing object detection accuracy. Our method excels at addressing the complexity of real-world scenes and diverse lighting conditions, making it highly effective across diverse and demanding environments.

Our hairstyle recommendation system integrates face shape classification, face swapping, and hair dyeing functionalities to aid customers in choosing the right style. Achieving 0.86 accuracy in face shape classification using VGG16, and a dice score of 0.98 with face swapping and U-Net, our system provides reliable references for hairstyle and color choices. It improves communication between stylists and customers, enhances design outcomes, and boosts satisfaction, contributing to the advancement of the hairdressing industry.

The task of image dehazing (DH) is an indispensable part of Advanced Driver Assistance Systems. This study proposes a method that combines image super- resolution (SR) and DH, focusing on how the models for DH and SR can be combined to reduce the overall computational cost. we added a modified version of SRCNN proposed in this study to help reducing the overall DH computations. In addition, a joint network training strategy for DH/SR combination is proposed. The experimental results show that the proposed method can reduce the computational complexity by about 70% compared to the original DH network – GridDehazeNet, while effectively preserving the image quality and clarity.

2023/4

Image recognition technology is pivotal in advanced driver assistance systems (ADAS). This study explores heterogeneous image fusion to enhance object detection in ADAS, focusing on the combination of infrared (IR) and visible (VIS) images due to their complementary nature. Addressing the alignment issue often overlooked in fusion studies, we review alignment and fusion methods, proposing an integrated approach for ADAS. Utilizing deep learning networks for pedestrian and vehicle detection pre- and post-fusion, our experiments demonstrate improved performance. Our approach outperforms prior fusion studies in both accuracy enhancement and execution speed. Importantly, we advocate using fused images directly for training deep learning networks to optimize object detection accuracy.