基于多特征融合的遥感图像特征提取方法
|
摘要
在进行遥感图像特征提取时,由于图像分辨率发生多维变换,使得图像特征多且不明显,传统的提取方法只考虑两个图像特征变量之间的关联性,没有考虑图像特征维数间的关联,导致图像特征提取不准确的问题。为解决上述问题,提出SVM和多特征融合的遥感图像特征提取方法,将SVM和DS证据理论相结合,在对遥感图像进行预处理后提取图像特征并进行分类,在将分类结果为依据,构造基本概率指派,引入利用矩阵分析的DS融合算法简化决策级融合算法复杂度;利用特征接近度的多特征融合算法,将特征矢量与类原图形模式之间的吸引力组成接近度矢量作为融合特征完成遥感图像特征的提取。仿真结果表明,上述方法在图像提取过程中,对像素的精准度和像素提取速度都比传统方法有很大提高,具有较强的可行性。
In the paper,a method of remote sensing image feature extraction was proposed based on SVM and multi- feature fusion. SVM and DS evidence theory were combined to make image feature extraction and classification after preprocessing of remote sensing image. Then the classification results were taken as the basis,to construct the basic probability assignment. DS fusion algorithm based on matrix analysis was introduced,to simplify the complexity of decision level fusion algorithm. Multi- feature fusion algorithm based on feature proximity was used to define the attractiveness and to express the spatial proximity between the characteristic vector and the original pattern.The attractiveness between characteristic vector and the original pattern was used to form the close degree vector,which is used as the fusion feature to complete the feature extraction of the remote sensing image. The simulation results show that the accuracy of pixel and speed of extraction pixel of the proposed method can be improved compared with the traditional method in the process of image extraction,which is high feasibility.
In the paper,a method of remote sensing image feature extraction was proposed based on SVM and multi- feature fusion. SVM and DS evidence theory were combined to make image feature extraction and classification after preprocessing of remote sensing image. Then the classification results were taken as the basis,to construct the basic probability assignment. DS fusion algorithm based on matrix analysis was introduced,to simplify the complexity of decision level fusion algorithm. Multi- feature fusion algorithm based on feature proximity was used to define the attractiveness and to express the spatial proximity between the characteristic vector and the original pattern.The attractiveness between characteristic vector and the original pattern was used to form the close degree vector,which is used as the fusion feature to complete the feature extraction of the remote sensing image. The simulation results show that the accuracy of pixel and speed of extraction pixel of the proposed method can be improved compared with the traditional method in the process of image extraction,which is high feasibility.
引文
[1]查易艺,孙权森,罗楠,等.基于FAST和DAISY的遥感图像配准算法[J].计算机应用研究,2016,33(2):624-628.
[2]冈甄勇,刘伟,陈建宏,等.高分辨率SAR图像舰船目标几何结构特征提取[J].信号处理,2016,32(4):424-429.
[3]杨磊,刘尚争.基于图像欧式距离和拉普拉斯特征映射的端元提取算法[J].电光与控制,2016(4):48-52.
[4]李建飞,文志强,胡永祥,等.基于改进区域生长的遥感影像道路提取[J].计算机工程与应用,2016,52(2):209-213.
[5]杨宜菩,杨帆,潘国峰,等.基于同态系统滤波的高分辨率遥感图像河流信息提取[J].计算机应用,2016,36(1):248-253.
[6]梁若飞,杨风暴,王毅敏,等.一种基于块模糊增强的玉米田遥感图像边缘检测方法[J].影像科学与光化学,2016,34(2):190-197.
[7]刘琼,严由嵘,张旭,等.基于动态小波亚像素配准的战场遥感识别仿真[J].计算机仿真,2015,32(9):1-4.
[8]李刚,吴增印.基于多尺度融合的高分辨率遥感图像分类[J].电子设计工程,2016,24(9):180-182.
[9]章志,陈艳拢,罗锋.基于遥感技术的2014年南黄海浒苔时空分布特征研究[J].淮海工学院学报:自然科学版,2016,25(1):80-85.
[10]侍昊,李旭文,牛志春,等.基于随机森林模型的太湖水生植被遥感信息提取[J].湖泊科学,2016,28(3):635-644.
[2]冈甄勇,刘伟,陈建宏,等.高分辨率SAR图像舰船目标几何结构特征提取[J].信号处理,2016,32(4):424-429.
[3]杨磊,刘尚争.基于图像欧式距离和拉普拉斯特征映射的端元提取算法[J].电光与控制,2016(4):48-52.
[4]李建飞,文志强,胡永祥,等.基于改进区域生长的遥感影像道路提取[J].计算机工程与应用,2016,52(2):209-213.
[5]杨宜菩,杨帆,潘国峰,等.基于同态系统滤波的高分辨率遥感图像河流信息提取[J].计算机应用,2016,36(1):248-253.
[6]梁若飞,杨风暴,王毅敏,等.一种基于块模糊增强的玉米田遥感图像边缘检测方法[J].影像科学与光化学,2016,34(2):190-197.
[7]刘琼,严由嵘,张旭,等.基于动态小波亚像素配准的战场遥感识别仿真[J].计算机仿真,2015,32(9):1-4.
[8]李刚,吴增印.基于多尺度融合的高分辨率遥感图像分类[J].电子设计工程,2016,24(9):180-182.
[9]章志,陈艳拢,罗锋.基于遥感技术的2014年南黄海浒苔时空分布特征研究[J].淮海工学院学报:自然科学版,2016,25(1):80-85.
[10]侍昊,李旭文,牛志春,等.基于随机森林模型的太湖水生植被遥感信息提取[J].湖泊科学,2016,28(3):635-644.