Maritime radar is the kernel sensor for tracking vessels in Vessel Traffic Service system, it is important for Maritime Situation Awareness. However, the images collected by the maritime radar are inundated with excessive noise blips, which bring variety of troubles in extraction of ship targets from the images. This paper proposes a radar target recognition method based on fuzzy C-means. First, the attributes of the blips in the sequential radar images, such as speed, direction and size, are quantified as three pieces of evidence to determine whether a radar blip is a moving vessel. Then, an artificial intelligence was built based on FCM. According to the three pieces of evidence, the possibility of a blip being a real vessel is computed with FCM. The main difficulty in building the FCM framework is to find an appropriate way to provide the classification coefficient C and the fuzzy coefficient m. Since the C in classification is finite, this study proposes a method to obtain C by assessing the Euclidean distance of the expected result. Since m is related to the discreteness of evidence and results, the coefficient can be assessed by Shannon entropy and gain. Field experiments suggest that the improved FCM is capable of classifying the radar blips accurately, and reducing the operational strength of the ship operators and improving the safety.