就图像单个运动目标识别和跟踪问题提出了一种解决方案。在图像分割部分,提出了基于数学形态学中流域分割的实现方法,依据单运动目标的特征通过合并过渡分割区域算法较好地解决了过度分割的现象,实现无需人工设置任何阈值,全智能化图像分割;在目标提取部分,提出了一种在连续多帧图像中自动提取单运动目标的方法,能适应摄像头固定和摄像头随运动目标移动等各种情况;在目标跟踪部分,对进行匹配的模板设置掩码,并自动进行调整,使得对运动目标的跟踪更为鲁棒。
This paper proposes an approach to recognize and track single moving object in image sequences.In the part of image segmentation,a method based on watershed segmentation in mathematical morphology is proposed,which could solve the oversegmentation problem,and achieve image segmentation intelligently,without any threshold value set manually.In the part of object capture,a method to capture single moving object automatically in image sequences is proposed,which could be used with both immovable camera and movable one.In the part of object tracking,a mask of the template is set to match the region,and is also adjusted in the process of tracking,which could lead to highly robust tracking.