GuardianLane – An Intelligent Road Safety System using Lane Detection, Road Sign Recognition, and SOS Response

Important tasks in Intelligent Transportation Systems (ITS) for autonomous driving include lane detection, traffic sign detection, and vehicle collision prediction. Different item sizes and weather conditions make it difficult to detect traffic signs. To improve detection accuracy, we suggest a modified YOLOv5s-based small object detection algorithm that uses Alpha-IoU for refined bounding box regression, an extra prediction head for fine-grained small object recognition, and coordinate attention for better feature extraction. Using the CCTSDB 2021 dataset, this model achieves 88.1% precision and 79.8% recall, greatly enhancing small-item detection in various weather scenarios. Cloud-edge computing also helps lane detection by lowering the computational burden and enhancing real-time performance. A CNN-based dual model that makes use of distributed computing and instance segmentation improves efficiency and guarantees precise lane recognition even in situations where the slope changes. Moreover, motion temporal templates and fuzzy time-slicing are used in vehicle accident prediction to follow moving objects and identify unusual driving patterns. Real-time monitoring capabilities are ensured by a deep neural network (DNN) trained on extracted features, which predicts accidents with a 98.5% hit rate and a 4.2% false alarm rate. Together, these developments enhance traffic monitoring, autonomous navigation, and road safety in smart cities.