Bridges are a crucial component of urban transportation systems, and their safety is directly linked to the life safety of road users. Real-time vibration and frequency monitoring of bridge structures can help identify potential issues at an early stage. By analyzing natural frequencies and vibration modes of bridges, early warnings for damage detection can be provided. While electronic accelerometers are commonly used, they are point-based and only offer localized measurements. Fiber optic strain sensors have recently attracted attention for vibration mode monitoring of bridges. To address challenges associated with high monitoring costs, complex sensor deployment, and intricate construction when using fiber optic sensors for bridge vibration mode monitoring, a monitoring system is proposed based on a localized array of Fiber Bragg Grating (FBG) strain sensors for the online vibration mode monitoring of medium- and small-span bridges. A high-resolution, long-range FBG strain gauge array is designed, consisting of a single optical fiber embedded with FBG sensors. The fiber is coated with a Glass Fiber Reinforced Polymer (GFRP) layer for protection and structural reinforcement. The sensor array contains seven monitoring points, each equipped with an FBG strain sensor. The sensors are mounted using a Fiber Bragg Grating (FBG) -based platform, which facilitates both sensor fixation and prestress adjustment. A Gray Wolf Optimization (GWO) -based Variational Mode Decomposition (VMD) method is introduced for the extraction of bridge modal parameters, including frequency, damping ratio, and mode shape. Through bridge model simulation experiments, the FBG strain gauge array is installed using a bonding technique. After extracting the bridge's resonance signals via the GWO-VMD method, a secondary interpolation process is employed to fit and extract strain values corresponding to the peak and trough points for the first-order modal analysis. The traditional Stochastic Subspace Identification (SSI) method is then utilized for first-order modal extraction, and the results from both methods are compared. The Pearson correlation coefficient between the two methods is 0.943 75, highlighting the effectiveness of the globally installed FBG strain gauge array in combination with the GWO-VMD method for modal parameter extraction. For practical validation, the system was tested on the Nansongshuigang Bridge along the national key highway from Rizhao to Nanyang. The FBG strain gauge array was positioned at the midspan of the bridge, with four sensor arrays evenly distributed along the bridge's underside. The results demonstrate that the designed FBG strain gauge array effectively captures strain signals induced by passing vehicles, achieving a dynamic strain resolution of 0.1 με. The GWO-VMD method successfully isolates the non-stationary vehicle-induced signals from the bridge's resonance signals. The second Intrinsic Mode Function (IMF2), which encapsulates the bridge's modal behavior, is used for modal analysis. Peak-trough analysis of IMF2 reveals a curve characterized by an initial increase followed by a decrease, with the peak corresponding to the midspan of the bridge and the trough exhibiting the opposite trend. Normalization of the results indicates that the first mode shape derived from the proposed method has a Pearson correlation coefficient of 0.95248 with the finite element simulation results. When compared to traditional electronic sensor monitoring methods, the proposed FBG strain gauge array demonstrates a first-order modal frequency error of less than 1% and a damping ratio error of less than 5% under vehicle excitation. The bridge's natural frequency exhibits a pattern of initial decrease followed by an increase, while temperature variations show an opposite trend. Additionally, the FBG strain gauge array is sensitive to small variations in the bridge's peak frequency due to temperature changes. The online monitoring system and methodology presented in this study enable the extraction of key bridge modal information-such as natural frequency, mode shape, and damping ratio-using a minimal number of strategically placed sensors. This approach facilitates real-time monitoring of the bridge's dynamic characteristics with low-cost implementation, providing a novel approach for assessing the structural health of bridges.