序列图像中运动目标的自动提取方法

针对目标检测与跟踪领域中的运动目标自动提取问题,提出了一种新的运动目标自动提取方法.利用已有的图像帧滤波后初始化背景,并在运动目标检测过程中,利用检测结果,不断地自动更新背景.使用背景差法检测运动区域,并对差分图像进行动态阈值分割,以及边缘链接,使其边缘处于基本连续状态.在得到的二值图上,提取轮廓,并根据目标大小选择面积阈值,剔除由于噪音或者背景提取不干净造成的虚假轮廓,将得到的轮廓掩模图像与原图像做逻辑与运算,提取出目标.实验结果表明,该方法可以有效地提取出刚体或非刚体运动目标.

Moving Object Auto-extraction in Image Sequences

A new method for moving object auto-extraction is presented in the field of moving object detection and tracking.The background model is initialized using filtered image frames,which have been stored,and the background is updated continuously by the result of object detection in the process of moving object detection.The moving region is obtained by the difference of current frame and the background.Then the adaptive threshold is used to get binary image and link the edge.The outlines of moving regions in binary image are extracted,and false outlines are eliminated resulted from noise and imprecise background by enacting area threshold value according to the size of moving object.Boolean calculation is done to the original image with the mask image including outlines of moving region and the result of extraction is obtained.The experiment results show that the method can well extract rigid body and nonrigid body can be extracted well by the presented method.