CAJ | 학술논문

针对高分辨率遥感图像近岸舰船的灰度与纹理特征与港口极其相似而较难实现舰船自动检测的问题，提出了一种基于活动基模型的近岸舰船检测算法。首先，学习船头的活动基模型来表示船头先验；然后，通过模板匹配来检测船头；最后，通过分析船头对称性来确定舰船主方向，并利用区域生长与上下文信息定位船身。实验结果表明，该方法不仅能够有效检测近岸舰船，而且对于舰船有附属物干扰的情况下，仍能取得好的检测效果。
침대고분변솔요감도상근안함선적회도여문리특정여항구겁기상사이교난실현함선자동검측적문제，제출료일충기우활동기모형적근안함선검측산법。수선，학습선두적활동기모형래표시선두선험；연후，통과모판필배래검측선두；최후，통과분석선두대칭성래학정함선주방향，병이용구역생장여상하문신식정위선신。실험결과표명，해방법불부능구유효검측근안함선，이차대우함선유부속물간우적정황하，잉능취득호적검측효과。
The gray‐scale and texture characteristics of ships ,anchoring in harbor area ,are extremely similar to that of the coast in high‐resolution remote sensing images .Thus it is difficult to detect the ships automatically .An inshore ship detection method which based on active basis model is proposed .First ,the ship bow’s prior information is expressed by learning active basis model of the ship head .T hen ,the head is detected by matching the template .Finally ,the main di‐rection of the ship is gained by analyzing the symmetry of ship heads .And the ship body is located by region growing and the full utility of the context information .The experiments show our method is able to detect many kinds of ships effec‐tively ,even in the complex situation such as disturbances of attachment .