摘要
根据遥感图像飞机目标的特点,提出一种基于不变性特征的支持向量机(SVM)识别算法。首先结合小波分解进行平移、旋转、缩放不变性特征提取;然后对基于遗传算法(GA)的SVM模型参数选择方法在核函数的选择、搜索空间的确定等方面进行改进,并用改进后的算法实现SVM模型参数选择。对480幅遥感图像进行仿真实验,得到97.56%的正确识别率。与BP神经网络相比,识别率高,验证了算法的有效性。
According to the characteristics of aircraft in remote sensing images ,this paper presents a novel method which applies Support Vector Machines(SVM) based on invariable features to recognize aircraft types. Firstly,accompanied by wavelet transform a novel translation,rotation and scale invariant feature extraction method is proposed. Secondly the SVM model parameters selection algorithm based on GA is improved on two aspects which are the kernel function selection and the research space constraint. Then the improved algorithm is used to select the SVM model parameters. Experiments are performed on 480 remote sensing images and a recognition rate of 97.56% is achieved. The method is higher in recognition rate compared with BP neural network. The experimental results show the validity of the algorithm.
出处
《现代电子技术》
2007年第12期115-118,126,共5页
Modern Electronics Technique
关键词
不变性特征提取
支持向量机
遗传算法
目标识别
遥感图像
invariable features extraction
support vector machines
genetic algorithm
target recognition
remote sensing images