复杂码头环境下的船舶检测与跟踪算法

运动模板算法在复杂环境下无法准确提取运动目标区域,并且依赖帧间间隔的选取,无法对减速运动目标取得良好检测效果。针对该缺点,提出了一种改进的运动模版算法。首先,对输入的视频序列采用Canny算子结合轮廓信息提取水岸边界线;然后,将运动历史图沿着水岸边界线进行水岸分离,消除岸上运动目标的干扰;接着,对水面区域进行形态学处理,消除背景中水面上非目标运动对象;最后,对形态学处理后的结果进行船舶轮廓检测,计算最大轮廓外接矩形的宽和高,结合船舶当前位置的尾部坐标重建船舶轮廓外接矩形,以此实现实时的、高准确度的船舶检测与跟踪。实验结果表明,在复杂水面环境下,该方法能够实现实时、准确的船舶目标检测与跟踪。

A ship detection and tracking algorithm under complex wharf environment

The motion template algorithm in complex environment cannot extract the region of moving objects accurately, and relying on the selection of inter frame spacing it cannot achieve good effect when detecting slow moving target. We propose an improved motion template algorithm. Firstly, we adopt the Canny operator as well as the contour information to extract the waterfront boundary line in the input video sequences. Secondly, we divide the motion history image (MHI) into two parts along the waterfront boundary line so as to eliminate the interference of the moving target on the water’s edge. Then, morphological manipulations on water area of the MHI are used to eliminate non-target objects in the background. Finally, we obtain the maximum width and height of the external rectangle of the contour by detecting the ship contour from the processed motion history image. The external rectangle of the contour is reconstructed according to the current obtained coordinates of the ship. Experimental results show that our method can achieve accurate and real-time ship detection and tracking under complex water environment.