This research presents a novel approach for real-time stress state estimation in steel bridges using Fiber Bragg Grating (FBG) sensors and image recognition techniques. The methodology involves creating a digital model of the bridge, comprising a global finite element model (FEM) and detailed sub-models of critical areas. A database of precomputed load cases is generated, and real-time sensor data is matched to this database using the developed fingerprinting method. Image recognition is employed to detect multiple load scenarios, enhancing the accuracy of stress estimations and ensuring linear scalability for multi-load situations. The accuracy of the developed model was tested using a scaled setup using a 3 meter long aluminium bridge, proving its effectiveness in real-world conditions. The results demonstrate the feasibility of this approach, with reasonable accuracy achieved in both single and multi-load scenarios. Future work should focus on improving model accuracy, enhancing image recognition algorithms, and optimizing computational performance for large-scale applications.