Key points are not available for this paper at this time.
The paper presents a vessel target detection algorithm to achieve the maritime visual surveillance, which aims to reduce the influence of clutter that exists in the background and improve the reliability of ship target detection. In the proposed detector, the main steps including background modeling, model training and updating and the segmentation of foreground, are all based on Guassian Mixture Model (GMM). By exploiting the characteristics of GMM, we simply determine whether new pixels, in the video, belong to the foreground. Having modeled surveillance region, we roll out the moving ship detection using background subtraction, segmenting the ship target by the continuity of the adjacent frames. The target precision rate of the algorithm is 97.29% and the false alarm probability is 22.83% in the experiments. Comparing with other algorithms, the results show that this algorithm can not only improve target precision rate, but also reduce false alarm probability, and greatly overcome the influence of large amount of clutter on the detection of moving ship objects in video background, effectively restraining the influence of the noise from the dynamic scenario transformation.
Chen et al. (Sat,) studied this question.