在基于视觉检测方式的泊位自动引导系统中,从序列图像中提取泊位飞机,检测泊位飞机的阴影区域,是泊位系统实现的关键。基于高斯混合模型的背景分割算法被广泛应用于静态背景分割中,但是该算法在处理高分辨率图像时,算法实时性显著下降 分割体积大而且运动缓慢的物体时,容易产生“拖尾”现象 不能检测出运动物体的阴影区域。为此提出了基于分层图像的改进高斯混合模型背景分割算法,有效地克服了算法实时性差和“拖尾”现象。在此基础上,提出了基于色彩特征和区域特征相结合的阴影检测算法,利用部分空间约束信息,检测出运动物体的阴影区域。实验结果表明了该算法的有效性和实用性。

-In based on the vision detection mode s berth homing guidance system, withdraws the berth airplane from the sequence image, examines the berth airplane s shadow region, is key which the berth system realizes. Is widely applied based on the Gauss mixed model s background division algorithm in the static background division, but this algorithm when processes the high resolution image, algorithm timeliness obviously drops　Division volume greatly when movement slow object, easy to have “trailing” the phenomenon　Cannot examine the mobile the shadow region. Therefore proposed improves the Gauss mixed model background division algorithm based on the lamination image, has overcome the algorithm timeliness effectively bad and “trailing” the phenomenon. Based on this, proposed unifies the shadow examination algorithm based on the color characteristic and the provincial characteristics, the use partial space restraint information, examines the mobile the shadow region. The experimental result h