基于分形理论的目标检测  

Fractal-based Image Target Detection

作　　者：赵巨波[1,2] 万建伟[1] 王永杰[2] 

机构地区：[1]国防科技大学电子科学与工程学院,湖南长沙410073 [2]中国人民解放军92941部队,辽宁葫芦岛125000

出　　处：《电子学报》2005年第B12期2524-2526,共3页Acta Electronica Sinica

摘　　要：军事目标检测一般是由自然背景和少量的人造目标所组成．现有的图像目标检测方法大多是基于对目标本身特性进行分析，不适用于目标特性未知的军事目标检测．本文根据人造目标和自然背景在分形特征上的固有差异，在利用分形维数提取人造目标边缘特征的基础上，结合小波分析方法提取的目标边缘特征，对自然背景中的人造军事目标进行检测．实验证明该方法具有较好的适应性．Generally,the image of ordnance target detection consists of nature background and a little artificial target. Most of the methods of image target detection now available are algorithms of analyzing the peculiarity of target itself,and do not apply in the ordnance target detection under the circumstances that the peculiarity of target is uncertain. A method of detecting artificial targets edges in natural background is presented. According to the intrinsic difference in fractal features between natural background and artificial objects, the method gets the edge feature of artificial target using fractal dimension and wavelet analysis. The effect of detection method is applicable by the results of experiments.

关 键 词：分形维数 小波分析 边缘特征 目标检测 

分 类 号：TN958[电子电信—信号与信息处理]