Traffic light recognition is a critical component of autonomous driving and intelligent transportation systems, requiring accurate and efficient detection in real-time scenarios. In this study, BiFPN-YOLOv8 is used as a high-performance deep learning model for traffic light detection, which enhances YOLOv8n by incorporating the Bidirectional Feature Pyramid Network (BiFPN) and an additional detection head. While YOLOv8n is known for its speed and accuracy, its ability to detect small and distant objects is limited. By integrating BiFPN, the model improves multi-scale feature fusion, leading to enhanced detection of small traffic lights under various lighting and occlusion conditions. Additionally, the inclusion of a fourth detection head (P2 layer) further strengthens the recognition of distant traffic signals. We evaluate the proposed model using the LISA dataset, consisting of 14,034 daylight traffic light images across six distinct classes. Experimental results demonstrate that BiFPN-YOLOv8n achieves a mean average precision (mAP) of 94.19%, outperforming baseline models such as YOLOv8n (92.31%) and Tiny-YOLOv7-3L (90.35%). These improvements make BiFPN-YOLOv8 a promising solution for real-world traffic light detection and other vision tasks requiring precise multi-scale object recognition.

Emerging indoor visible light communication and positioning (VLCP) technology is placed among those of recent hot topics as it is capable of declaring the spatial location of objects, providing users with highly accurate positioning coordinates in 3-D with side benefits e.g., no RF interference generation, low cost, and minimized hardware. The current demand for multiple applications in various fields has called for the development of a robust, accurate, and optimized device that leverages the capabilities of VLCP technology. This system is of particular interest as it promises to deliver reliable results for numerous applications. In this paper, the design procedure of high precision indoor 3-D VLCP system is defined as an optimization problem which is solved by utilizing metaheuristic algorithms.