利用支持向量机的飞机目标检测.pdfVIP

15 9 V l. 15 N . 9 2008 9 Electr nics Optics C ntr l Sep. 2008 : 1671- 637 Ú( 2008) 09-0006-04 黄 洁, 张 海 ( , 100083) : 为了在有干扰弹和飞机机型不定的情况下检测出天空背景中的飞机, 通过自动阈值 割和腐蚀膨胀运算找出图像中的疑似区, 提取其灰度方差不变矩特征和像素比组成特征向量, 应用支持向量机作为 类器对疑似区进行判别, 判断其是否为飞机, 选取典型的飞机图像作为正样 本, 选取干扰弹图像作为负样本用不同背景和飞机机型的图像进行试验, 证明该方法能很好地提 出疑似区并有效地对疑似区进行检测识别出飞机目标 : 目标检测; 阈值 割; 不变矩; 支持向量机; 飞机识别 : V271. 4; TP391. 41 : A Plane detection based on support vector machine and information fusion HUANG Jie, ZHANG Hai (AutomationScience and Electrical Engineering College, Bejiing University of Aeronautics and Astronautics, Beij ing 100083, China) Abstract : In rder t detect planes with ut fixed appearances under sky backgr und when interfering b mbs existed, aut matic thresh lding meth d and er si n/ dilati n arithmetic are applied t find ut the suspici us areas f aircraft. The gray variance, invariant m ments and the rate f number f pixel are extracted t c nstitute the image feature vect r. Then the meth d f Supp rt Vect r Machine ( SVM) is used as a classifier t detect the suspici us areas and t judge if it is a plane r n t. S me typical aircrafts are ch sen as the p sitive samples and s me interfering b mbs are ch sen as the negative samples. Using images f different backgr unds and different aircraft appearances as test images, it is dem nstrated that this meth d is effective n suspici us areas search and aircraft rec gniti n. Key words: bject detecti n; thresh lding; invariant m ments; supp rt vect r machine; plane rec gniti n