基于帧差法和混合高斯的海上运动目标检测  

A Method for Detecting Maritime Moving Targets Based on Three-frame Difference Method and Improved Hybrid Gaussian Background Model

作　　者：高海壮 段先华 GAO Haizhuang;DUAN Xianhua(School of Computer Science and Engineering,Jiangsu University of Science and Technology,Zhenjiang 212000)

机构地区：[1]江苏科技大学计算机学院,镇江212000

出　　处：《计算机与数字工程》2019年第5期1140-1144,共5页Computer & Digital Engineering

摘　　要：海上运动目标检测是视频信息处理的重要研究内容之一。由于海上场景的复杂性,传统的高斯混合背景模型存在一定的不足。因此论文提出了一种三帧差分法和改进的混合高斯背景模型相结合的运动目标检测算法。首先,用三帧差分法检测出变化的区域;然后,根据相邻像素间的相关性,把每帧图像划分为背景区域、干扰区域和目标区域,对不同的区域采取不同的更新策略,从而实现模型的自适应学习。实验表明,该方法可以有效地抑制海上波纹对运动目标检测的影响,并具有较好的灵敏度、自适应性和良好的检测效果。The detection of marine moving target is one of the important research contents of video information processing. Due to the complexity of the sea scene,the traditional Gaussian mixed background model has some shortcomings. Therefore,this paper presents a three-frame difference method and an improved hybrid Gaussian background model combined with the moving target de. tection algorithm. Firstly,the change of the region is detected by the three - frame difference method. Then,according to the corre. lation between adjacent pixels,each frame image is divided into background region,interference region and target area,and differ. ent updating strategies are adopted for different regions,so as to realize the adaptive learning of the model. Experiments show that the method can effectively suppress the influence of sea ripple on the detection of moving objects,and has good sensitivity, self-adaptability and good detection results.

关 键 词：海上运动目标检测 混合高斯背景模型 三帧差分法 更新策略 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]