Single Shot Multibox detector based video surveillance system for active anti-collision between vessel and bridge is proposed aiming at the limitations of the existing systems in accuracy，robustness and efficiency. A vessel-exclusive dataset with tons of image samples is established for neural network training，and an SSD (Single Shot MultiBox Detector) based object detection model with both universality and pertinence is generated with tactics of sample filtering，data augmentation and large-scale optimization，which can realize the stable and intelligent vessel detection. Comparison results with conventional methods indicate that the proposed navigational object detection method shows remarkable advantage in robustness，accuracy，efficiency and intelligence. In-situ test is carried out at Songpu Bridge in Shanghai，and the results illustrate that the method is qualified for long-term monitoring and provides information support for further analysis and decision making.