基于meanshift滤波和蚁群聚类的复杂红外目标分割  

Complex Infrared Objects Segmentation Based on Mean-shift Filtering and Ant Colony Clustering

作　　者：王潇晨[1] 付冬梅[1] 李晓刚[1] 刘燕[2] 

机构地区：[1]北京科技大学,北京100083 [2]中国人民解放军装甲兵工程学院,北京100072

出　　处：《红外技术》2009年第9期545-548,共4页Infrared Technology

基　　金：国家自然科学基金资助项目(编号:60573016)

摘　　要：提出了一种复杂背景下红外目标分割的有效方法。该方法首先利用meanshift的自适应平滑滤波特性,在不损失目标信息的情况下,滤除复杂背景的杂波干扰;然后根据滤波得到的区域,用蚁群方法在区域之间进行聚类合并,获得最终的分割结果。采用区域来表征蚂蚁,与基本蚁群算法将每个像素看作一只蚂蚁相比,其蚂蚁个数大大减少,因而减小了计算的复杂度,提高了图像处理的效率。在蚁群算法中引入了一种新的引导函数,可以更准确引导蚁群聚类。实验结果表明,该方法可以准确地分割出目标,是一种快速有效的图像分割方法。A novel algorithm was presented to achieve an improved complex infrared objects segmentation performance. First, the algorithm uses discontinuity preserving smoothing algorithm based on mean-shift procedure to filter the powerful noise without the loss of the object information. Second, the regions produced by mean-shift filtering were merged by ant colony clustering algorithm to gain result of image segmentation. Due to the less ants produced by the regio~ s of filtered image than the original image, and a new visibility based on intensity distribution is defined, it＇s more accurate and efficient to cluster ant colony. Experimental results show the superior performance of the proposed infrared object segmentation algorithm.

关 键 词：复杂红外目标 MEANSHIFT 蚁群聚类 图像分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]