Laura Martinho

Driver assistance techniques are increasingly required to operate in real-world environments with low cost and limited computational resources. However, most existing schemes rely on multi-sensor configurations or complex calibration procedures, which increase system cost and limit scalability. This paper presents a lightweight monocular methodology for real-time risk analysis, designed for embedded automotive applications. It integrates a homography-based calibration procedure, bird’s eye view transformation, and distance estimation to obtain spatial information from a single camera. Based on this representation, it computes proximity to critical regions and assigns risk levels using a distance-based approach, avoiding dependency on predefined polygonal zones. Experimental observations show that the proposed approach provides consistent risk estimation and stable behavior across different scenarios, while maintaining low computational cost and real-time performance.