Abstract: A non-contact identification system of bridge deflection based on network cameras is established to solve the problem of poor accessibility, of low efficiency, and of incompetence in all-weather real-time monitoring of the traditional bridge deflection identification system. The system used LED (light-emitting diode) lights as the marker and took network cameras as the acquisition equipment to transmit the image information wirelessly, the deflection time history information of the target is obtained by using the computer carries HSV (hue, saturation, and value) based on fast template matching and color tracking (cvCamShift) based on geometric matching algorithm, and then the non-contact identification of bridge deflection can be realized. The feasibility of the system is verified by the vibration tests under four working conditions on the footbridge model and the field test of a real bridge. The results show that: in the model test, the error of the time domain and of the frequency domain information identified by the system compared with the laser displacement sensor is less than 0.6%, and the error can still be less than 0.7% under the interference of fog. Under the real bridge test, the error of the system identification results is less than 1.9% compared with the bridge deflection instrument. This indicates that the system has strong robustness and good economy and has a wide application prospect.