摘要
针对当前图像目标分类识别方法识别率低、耗时较长等问题,提出基于支持向量机的非接触数字图像注入式红外目标分类识别方法。采用尺度不变特征转换算法对图像注入式红外目标进行特征提取,提取的结果含有全局信息的全局向量。根据SVM训练中支持向量的分布特点得到目标判别函数,通过分析判别函数与SVM分类器性能之间的关系,选择组合判别函数构造SVM分类器,将注入式红外目标输入到SVM分类器中进行目标分类识别,完成对数字图像注入式红外目标分类识别。仿真证明,所提方法不仅能够更精确地识别和提取图像特征,并通过降低维数复杂度减少了识别时间。
At present,the method has low recognition rate and long time consumption.Therefore,we focus on a method to classify and recognize injection infrared target in non-contact digital image based on support.vector machine.Firstly,we used the scale-invariant feature transformation algorithm to extract feature from the injected infrared target of image.The extracted result contained the global vector of global information.According to the distribu-tion characteristic of support vector in SVM training,we obtained the target discriminant function.By analyzing the relation between discriminant function and performance of SVM classifier,we chose compound discriminant function to construct SVM classifier.Flnally,we inputted the injected infrared target to SVM classifier for target for the classification recognition.Thus,we completed the classification recognition of injection infrared target in digital image. Simulation proves that the proposed method can not only recognize and extract image feature accurately,but also re- duce recognition time by reducing the complexity degree of dimension.
作者
宁海涛
NING Hai-tao(College of Humanities & Information,Changchun University of Technology,Changchun Jilin 130122,China)
出处
《计算机仿真》
北大核心
2018年第12期376-379,共4页
Computer Simulation
关键词
非接触
数字图像
注入式
红外目标
分类识别
支持向量机
Non-contact
Digital image
Injection type
Infrared target
Classification recognition
Support vector machine (SVM)