一種基于Pauli分解和支持向量機的全極化合成孔徑雷達監(jiān)督分類算法
發(fā)布時間：2018-10-29 18:11
【摘要】：全極化合成孔徑雷達(SAR)影像準(zhǔn)確分類的一個重要前提是充分提取反映地物實際物理性質(zhì)的特征。然而現(xiàn)有的全極化SAR特征提取算法和分類算法眾多,卻均存在各種各樣的問題。無論極化特征提取方法還是分類算法,都會影響最終的分類精度。針對此問題,在多次實驗的基礎(chǔ)上,提出一種綜合Pauli極化特征分解和支持向量機(SVM)的分類策略,簡稱為Pauli-SVM算法。首先通過經(jīng)典的Pauli分解法提取全極化SAR影像的奇次散射、偶次散射、體散射等極化特征;并將這些信息組合成一個特征向量,然后引入高精度的SVM分類算法,選擇訓(xùn)練樣本后對全極化SAR影像進行監(jiān)督分類。在江蘇溧水和南京橫溪鎮(zhèn)兩個研究區(qū),以ALOS衛(wèi)星的PALSAR影像為研究數(shù)據(jù),進行監(jiān)督Wishart分類算法、Freeman特征提取法結(jié)合SVM的分類算法、Yamaguchi特征提取法結(jié)合SVM的分類算法、Pauli-SVM算法的分類對比實驗。結(jié)果表明,新提出的PauliSVM算法可以有效地提高分類的準(zhǔn)確性。
[Abstract]:An important prerequisite for accurate classification of fully polarized synthetic Aperture Radar (SAR) images is to fully extract features that reflect the actual physical properties of ground objects. However, there are many existing full-polarization SAR feature extraction algorithms and classification algorithms, but there are a variety of problems. Whether polarization feature extraction method or classification algorithm will affect the final classification accuracy. In order to solve this problem, on the basis of many experiments, a classification strategy based on Pauli polarization feature decomposition and support vector machine (SVM) is proposed, which is called Pauli-SVM algorithm for short. Firstly, the odd scattering, even scattering, volume scattering and other polarization characteristics of fully polarized SAR images are extracted by classical Pauli decomposition method. The information is combined into a feature vector, and then a high-precision SVM classification algorithm is introduced. The training samples are selected and supervised classification of fully polarized SAR images is carried out. In two research areas of Lishui, Jiangsu Province and Hengxi Town, Nanjing, the supervised Wishart classification algorithm was carried out based on the PALSAR images of ALOS satellite, the Freeman feature extraction method combined with the SVM classification algorithm, and the Yamaguchi feature extraction method combined with SVM classification algorithm. Comparison experiment of Pauli-SVM algorithm. The results show that the proposed PauliSVM algorithm can effectively improve the accuracy of classification.
【作者單位】： 江蘇省資源環(huán)境信息工程重點實驗室(中國礦業(yè)大學(xué));江蘇省地理信息技術(shù)重點實驗室;
【基金】：國家自然科學(xué)基金(41171323) 中國地質(zhì)調(diào)查局地質(zhì)調(diào)查工作項目(1212011120229) 江蘇省自然科學(xué)基金(BK2012018) 地理空間信息工程國家測繪地理信息局重點實驗室開放基金(201109)資助
【分類號】：TN957.52
本文編號：2298418
[Abstract]:An important prerequisite for accurate classification of fully polarized synthetic Aperture Radar (SAR) images is to fully extract features that reflect the actual physical properties of ground objects. However, there are many existing full-polarization SAR feature extraction algorithms and classification algorithms, but there are a variety of problems. Whether polarization feature extraction method or classification algorithm will affect the final classification accuracy. In order to solve this problem, on the basis of many experiments, a classification strategy based on Pauli polarization feature decomposition and support vector machine (SVM) is proposed, which is called Pauli-SVM algorithm for short. Firstly, the odd scattering, even scattering, volume scattering and other polarization characteristics of fully polarized SAR images are extracted by classical Pauli decomposition method. The information is combined into a feature vector, and then a high-precision SVM classification algorithm is introduced. The training samples are selected and supervised classification of fully polarized SAR images is carried out. In two research areas of Lishui, Jiangsu Province and Hengxi Town, Nanjing, the supervised Wishart classification algorithm was carried out based on the PALSAR images of ALOS satellite, the Freeman feature extraction method combined with the SVM classification algorithm, and the Yamaguchi feature extraction method combined with SVM classification algorithm. Comparison experiment of Pauli-SVM algorithm. The results show that the proposed PauliSVM algorithm can effectively improve the accuracy of classification.
【作者單位】： 江蘇省資源環(huán)境信息工程重點實驗室(中國礦業(yè)大學(xué));江蘇省地理信息技術(shù)重點實驗室;
【基金】：國家自然科學(xué)基金(41171323) 中國地質(zhì)調(diào)查局地質(zhì)調(diào)查工作項目(1212011120229) 江蘇省自然科學(xué)基金(BK2012018) 地理空間信息工程國家測繪地理信息局重點實驗室開放基金(201109)資助
【分類號】：TN957.52
【相似文獻】
相關(guān)會議論文 前1條
1 ;More Entanglement,Less Reliability[A];中國自動化學(xué)會控制理論專業(yè)委員會B卷[C];2011年
相關(guān)碩士學(xué)位論文 前3條
1 曹嘯宇;量子圖態(tài)糾纏以及穩(wěn)定字碼在Pauli信道中的性能[D];浙江工商大學(xué);2012年
2 劉超;量子非對稱碼用于振幅或相位衰減信道的性能研究[D];浙江工商大學(xué);2012年
3 王巖巖;量子穩(wěn)定子碼的量子容錯編碼門構(gòu)造方法研究[D];南京郵電大學(xué);2013年
，本文編號：2298418
本文鏈接：http://sikaile.net/kejilunwen/wltx/2298418.html
最近更新
教材專著
