Abstract: A feasible method combining with the character of marine visible images was proposed to detect targets automatically in the sequential images from surface vehicle. It is not only suitable for small targets under seasky background but also for large targets under offshore background. First, the complexity of subimages and the average gray difference of their up and down neighborhood are measured to predict the sea line region, canny edge detection with surround suppression is applied to extract the contour of the region and Hough transforming is used to pick the longest line as sea line. There is no more processing if the image does not possess sea line region. Second, the region is defined as simple seasky background or complex offshore background according to the variance above the sea line. Clustering is implemented to extract the targets within the local region tailored by the sea line based on its smooth change of gray level. Third, gray characteristic of the rest sea is analyzed and sea gray range is obtained. Thus sea blocks and target blocks can be distinguished quickly. Finally, the segmented image is done with proper postprocessing in order to discard pseudo targets and mark the real targets. The experimental results prove that this method can locate the targets exactly. The time cost per frame is within 100 ms. It is robust and fast.  

Key words: image processing, sea line range, surround suppression, clustering, sea gray range