基于壓縮感知的遙感影像彈性配準(zhǔn)方法
發(fā)布時間：2018-12-09 13:51
【摘要】：針對遙感影像由于載荷類型、觀測角度、地形起伏等內(nèi)外部因素造成的影像局部幾何畸變,而基于全局配準(zhǔn)方法制約著影像配準(zhǔn)精度的提高,基于像元的彈性配準(zhǔn)方法可大幅提升遙感影像的配準(zhǔn)精度,但是存在運(yùn)算效率這一瓶頸等問題,該文利用像元彈性配準(zhǔn)參數(shù)的稀疏性,提出一種基于壓縮感知的彈性配準(zhǔn)方法。通過對遙感影像像元梯度幅值響應(yīng)較強(qiáng)的點(diǎn)進(jìn)行隨機(jī)抽樣,形成觀測樣本點(diǎn)集,采用彈性配準(zhǔn)局部參數(shù)解算模型求解樣本點(diǎn)平移參數(shù);通過壓縮感知稀疏重構(gòu)算法重構(gòu)影像各像元平移參數(shù)。實(shí)驗表明,在配準(zhǔn)精度差異較小和一定的參數(shù)設(shè)置條件下,該方法可顯著提高彈性配準(zhǔn)運(yùn)算速度。
[Abstract]:Aiming at the local geometric distortion of remote sensing image caused by internal and external factors such as load type, observation angle, topography fluctuation and so on, the improvement of image registration accuracy is restricted by global registration method. The elastic registration method based on pixel can greatly improve the registration accuracy of remote sensing image, but there are some problems such as the bottleneck of computing efficiency. This paper proposes an elastic registration method based on compressed sensing based on the sparsity of pixel elastic registration parameters. Through random sampling of the points with strong gradient amplitude response of image pixel, the observation sample point set is formed, and the translation parameter of the sample point is solved by using the elastic registration local parameter solution model. The translational parameters of each pixel are reconstructed by compressed sparse reconstruction algorithm. The experimental results show that this method can significantly improve the speed of elastic registration under the condition of little difference in registration accuracy and certain parameter setting.
【作者單位】： 中國科學(xué)院電子學(xué)研究所;中國科學(xué)院光電研究院;中科九度(北京)空間信息技術(shù)有限責(zé)任公司;
【基金】：中國科學(xué)院科技服務(wù)網(wǎng)絡(luò)計劃項目(KFJ-EW-STS-046) 國家高技術(shù)研究發(fā)展計劃項目(2014AA09A511) 高分辨率對地觀測系統(tǒng)重大專項(E0303/1315/05)
【分類號】：TP751
本文編號：2369459
[Abstract]:Aiming at the local geometric distortion of remote sensing image caused by internal and external factors such as load type, observation angle, topography fluctuation and so on, the improvement of image registration accuracy is restricted by global registration method. The elastic registration method based on pixel can greatly improve the registration accuracy of remote sensing image, but there are some problems such as the bottleneck of computing efficiency. This paper proposes an elastic registration method based on compressed sensing based on the sparsity of pixel elastic registration parameters. Through random sampling of the points with strong gradient amplitude response of image pixel, the observation sample point set is formed, and the translation parameter of the sample point is solved by using the elastic registration local parameter solution model. The translational parameters of each pixel are reconstructed by compressed sparse reconstruction algorithm. The experimental results show that this method can significantly improve the speed of elastic registration under the condition of little difference in registration accuracy and certain parameter setting.
【作者單位】： 中國科學(xué)院電子學(xué)研究所;中國科學(xué)院光電研究院;中科九度(北京)空間信息技術(shù)有限責(zé)任公司;
【基金】：中國科學(xué)院科技服務(wù)網(wǎng)絡(luò)計劃項目(KFJ-EW-STS-046) 國家高技術(shù)研究發(fā)展計劃項目(2014AA09A511) 高分辨率對地觀測系統(tǒng)重大專項(E0303/1315/05)
【分類號】：TP751
【相似文獻(xiàn)】
相關(guān)期刊論文 前1條
1 張紅穎;張加萬;孫濟(jì)洲;楊甲東;;基于層次B樣條的醫(yī)學(xué)圖像彈性配準(zhǔn)方法[J];天津大學(xué)學(xué)報;2007年01期
相關(guān)碩士學(xué)位論文 前1條
1 孫亞蘭;基于改進(jìn)互信息的多尺度彈性配準(zhǔn)方法研究[D];湘潭大學(xué);2008年
，本文編號：2369459
本文鏈接：http://www.wukwdryxk.cn/guanlilunwen/gongchengguanli/2369459.html
最近更新
教材專著
