光学技术, 2018, 44 (3): 291, 网络出版: 2018-06-09  

联合矩阵低秩逼近和稀疏表示的高分辨率遥感影像目标识别方法

A target recognition method of high resolution remote sensing images based on joint matrix low rank approximation and sparse representation
孔艳 1,*王保云 1,2何苗 1
作者单位
1 云南师范大学 信息学院, 云南 昆明 650500
2 西部资源环境地理信息技术教育部工程研究中心, 云南 昆明 650500
摘要
针对高分辨率遥感影像地物信息复杂、目标识别率低等问题, 提出了一种联合矩阵低秩逼近的稀疏表示遥感影像目标识别方法。对原始遥感影像进行Radon变换, 将处理过后的遥感影像进行低秩和稀疏分解, 得到具有低秩性和稀疏性的两部分信息; 通过K-SVD算法分别对这两部分信息进行字典学习, 构建稀疏表示的判别字典; 通过稀疏表示求解算法求解出待分类的目标在判别字典上的稀疏系数, 根据稀疏系数最大准则对目标进行分类识别。在Uc Merced 数据集上选取具有代表性的线性和非线性子集分别进行实验, 结果表明所提算法与传统的SRC、SVM、MLC和KNN等分类识别算法相比, 在采样比例为1/16、稀疏度为5时, 识别率在线性子集上能够提高10%、在非线性子集上能够提高5%, 表明所提方法具有较好的识别效果。
Abstract
A new method of high resolution remote sensing image recognition for sparse representation of joint matrix low rank approximation is proposed to solve the problem that the high resolution remote sensing image is complex and the target recognition rate is low. The Radon transform of the original remote sensing image was carried out, and the remote sensing image was decomposed into low rank and sparse. The two parts were classified as low rank and sparseness. Through the K-SVD algorithm, the dictionaries of the parts were studied, which were combined to build the discrimination dictionary of sparse representation. Finally, Through the sparse representation algorithm, the sparse coefficient of the classification target in the discrimination dictionary were obtained, and the target according to the maximum criterion of the sparse coefficient were classified. Experiments were carried out by selecting representative linear and non-linear subsets on the Uc Merced dataset, and the experimental result show that the algorithm can improve the recognition rate by about 10% in the linear subset and by about 5% in the non-linear subset compared with the traditional classification and recognition algorithms such as SRC, SVM, MLC and KNN in 1/16 of the sampling ratio and 5 of the sparseness , which indicates that the proposed method has better recognition effect.

孔艳, 王保云, 何苗. 联合矩阵低秩逼近和稀疏表示的高分辨率遥感影像目标识别方法[J]. 光学技术, 2018, 44(3): 291. KONG Yan, WANG Baoyun, HE Miao. A target recognition method of high resolution remote sensing images based on joint matrix low rank approximation and sparse representation[J]. Optical Technique, 2018, 44(3): 291.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!