[Objectives] In order to meet the needs of long-distance sea target recognition and tracking through the navigation radar and the integrated optoelectronic when the unmanned ship operates in the marine environment, a novel unmanned boat autonomous sensing system with heterogeneous sensor association, target track prediction, and photoelectric camera attitude compensation is developed. [Methods] The coordinate system of each heterogeneous sensor is unified through the registration algorithm of sensors. As for the target position information output by the navigation radar, a target track prediction algorithm based on the Kalman filter is used to improve the accuracy of target location and provide better real-time information for photoelectric cameras. Finally, the posture compensation algorithm based on the ship posture in the photoelectric camera is used to complete the tasks of image collection, recognition and tracking for the target. The unmanned surface vehicle equipped with our proposed sensing system has completed the dynamic target recognition and tracking tasks in the third sea state. [Results]Specifically, the target tracking error is reduced by 6% and the target recognition success rate is increased to 96.25%, which verifies the environmental adaptability of the sensing system. [Conclusions] With test experiments, the recognition and tracking system can effectively solve the problems of difficult image acquisition and poor recognition effect of sea surface target, and can effectively improve the success rate of target recognition.