基于特征点和区域生长的目标图像分割方法

成像探测的运动目标图像中背景复杂并且含有大量的噪声,针对传统的目标的检测和分割方法精确定位困难、且不能完整分割等问题,提出基于特征点和区域生长的运动目标图像分割方法。通过相邻帧图像的绝对值差分图像得到大概的运动区域,利用基于LK光流的角点检测方法提取差值图像中的特征点,采用非最大值抑制对特征点的优劣性进行评估,对好的特征点进行区域生长,最终达到运动目标的分割目的。仿真结果表明:该方法能够对复杂图像序列中的运动目标进行精确定位,得到较好的目标分割结果,并且计算量小,具有较高的鲁棒性。

Target Image Segmentation Method Based on Feature Selection and Region Growing

According to the problems that precise locattion was hard and perfect segmentation was not available for moving target in imaging detection under complex background and a large amount of noises,a moving target detection and segmentaion method based on image feature selection and region growing method was proposed.The probable moving region was obtained through the absolute difference image,the feature points were detected in difference image by using LK optical flow based corner detection method.The advantages evaluation was performed for the feature points,which would be used for region growing,that would result in the movingtarget segmentation.Simulation results showed that the method for moving target detection under complex background was of precise location,perfect segmentation,low computational complexity and strong robusticity.