本文選題：SIFT算法 + RANSAC算法��； 參考：《山東農(nóng)業(yè)大學(xué)》2013年碩士論文

【摘要】：由于SAR影像特殊的成像機(jī)理，導(dǎo)致SAR影像匹配的成功率、正確率、精度及匹配效率較低，SAR影像匹配成為SAR影像應(yīng)用的技術(shù)難點(diǎn)之一。本文針對(duì)機(jī)載SAR影像特殊的成像特點(diǎn)，以機(jī)載SAR影像匹配算法流程為主線，對(duì)機(jī)載SAR影像SIFT匹配算法進(jìn)行研究，并引入了基于2D單應(yīng)變換的RANSAC剔除誤匹配點(diǎn)對(duì)算法及基于物方約束的匹配點(diǎn)預(yù)測(cè)算法，對(duì)SIFT匹配算法進(jìn)行改進(jìn)、總結(jié)；論文重點(diǎn)研究了SIFT匹配算法，并提出了兩種改進(jìn)SIFT匹配算法的方法，第一種是SIFT和粗差剔除算法相結(jié)合的匹配方法；第二種是基于物方約束的SIFT匹配方法。前者是利用SIFT算法提取特征穩(wěn)定的同名點(diǎn)對(duì)，并結(jié)合基于2D單應(yīng)變換的RANSAC算法剔除SIFT誤匹配點(diǎn)對(duì)，以提高匹配點(diǎn)對(duì)的精度。而后者對(duì)前者進(jìn)一步改進(jìn)，基于物方約束的SIFT匹配方法是針對(duì)機(jī)載SAR影像特殊的成像特點(diǎn)和幾何特點(diǎn)，將機(jī)載SAR影像的幾何約束加入到SIFT算法匹配中，提高SIFT匹配點(diǎn)對(duì)的準(zhǔn)確率、精度及效率。 首先，以機(jī)載SAR影像匹配SIFT算法匹配流程為基礎(chǔ)，通過(guò)機(jī)載SAR影像SIFT匹配實(shí)驗(yàn)，從選取的機(jī)載SAR影像中三組不同地物類別的代表區(qū)域角度來(lái)分析，驗(yàn)證了SIFT算法能夠較準(zhǔn)確地匹配到穩(wěn)定的特征，具有一定的魯棒性。然后，在機(jī)載SAR影像匹配算法流程分析基礎(chǔ)上，通過(guò)編程，對(duì)選取的機(jī)載SAR影像中三組不同地物類別的代表區(qū)域分別進(jìn)行了SIFT和粗差剔除算法相結(jié)合的匹配實(shí)驗(yàn)和基于物方約束的SIFT匹配實(shí)驗(yàn)，并對(duì)匹配點(diǎn)對(duì)結(jié)果精度進(jìn)行了分析和評(píng)價(jià)。 本文研究的內(nèi)容和創(chuàng)新點(diǎn)如下： （1）本文分別采用了三種不同地物類別的機(jī)載SAR數(shù)據(jù)做實(shí)驗(yàn)，在含有人工建筑物、含有自然植被及紋理信息缺乏的機(jī)載SAR影像中，，從不同角度說(shuō)明SIFT算法可有效地提取穩(wěn)定的匹配點(diǎn)對(duì)，其正確率高，即使是在紋理信息缺乏的區(qū)域，便于在實(shí)際中實(shí)現(xiàn)整景SAR影像間重疊區(qū)域的匹配。 （2）引入了2D單應(yīng)變換以作為RANSAC算法剔除SIFT誤匹配點(diǎn)對(duì)的模型，提出了利用RANSAC算法剔除SIFT誤匹配點(diǎn)對(duì)，在基于2D單應(yīng)變換的RANSAC算法機(jī)載SAR影像誤匹配剔除實(shí)驗(yàn)中，通過(guò)對(duì)匹配點(diǎn)對(duì)數(shù)據(jù)的分析，驗(yàn)證了該方法可以有效地剔除SIFT誤匹配點(diǎn)對(duì)，進(jìn)而提高了SIFT匹配點(diǎn)對(duì)的準(zhǔn)確率和精度。 （3）分析了合成孔徑雷達(dá)的成像機(jī)理、影像特點(diǎn)和構(gòu)像模型，闡明了SAR影像匹配必須考慮到其自身影像特點(diǎn)。針對(duì)機(jī)載SAR影像特殊的成像特點(diǎn)，提出了以R-D模型為基礎(chǔ)，POS與DEM數(shù)據(jù)相結(jié)合輔助像點(diǎn)定位，將機(jī)載SAR影像的幾何約束加入到SIFT算法匹配中，利用物方約束來(lái)預(yù)測(cè)待匹配SAR影像的匹配點(diǎn)，建立以預(yù)測(cè)匹配點(diǎn)為中心的匹配搜索窗口，利用SIFT算法在此約束范圍內(nèi)進(jìn)行匹配。 （4）基于物方幾何約束的SIFT匹配實(shí)驗(yàn)，與利用SIFT和粗差剔除相結(jié)合的匹配算法相比，大大減少了誤匹配點(diǎn)對(duì)，進(jìn)一步提高了匹配點(diǎn)對(duì)的準(zhǔn)確率和精度；基于物方幾何約束的SIFT匹配并且約束了待匹配SAR影像的搜索匹配范圍，進(jìn)而提高了匹配效率�；谖锓綆缀渭s束的SIFT匹配實(shí)驗(yàn)結(jié)果表明該方法是一種非常有效的機(jī)載SAR影像匹配算法。
[Abstract]:Due to the special imaging mechanism of SAR image, the success rate, accuracy, accuracy and matching efficiency of SAR image matching are low. SAR image matching has become one of the technical difficulties in the application of SAR images. This paper, aiming at the special imaging characteristics of airborne SAR images, takes the process of airborne SAR image matching arithmetic as the main line, and advances the algorithm of SIFT matching for airborne SAR images. In this paper, the algorithm of RANSAC elimination mismatch point pair based on 2D single transformation and the matching point prediction algorithm based on object constraint are introduced, and the SIFT matching algorithm is improved and summed up. The paper focuses on the SIFT matching algorithm, and two methods to improve the SIFT matching algorithm are proposed. The first is the combination of SIFT and the coarse difference elimination algorithm. The second is the SIFT matching method based on the object constraint. The former uses the SIFT algorithm to extract the homonym pairs of the stable feature, and combines the RANSAC algorithm based on the 2D single stress transform to eliminate the SIFT mismatch point pair, in order to improve the accuracy of the matching point pair. The latter improves the former one step and the SIFT matching method based on the object constraint. In view of the special imaging features and geometric features of the airborne SAR image, the geometric constraints of the airborne SAR image are added to the SIFT algorithm matching to improve the accuracy, accuracy and efficiency of the matching point pairs of the SIFT.
First, on the basis of the matching process of airborne SAR image matching SIFT algorithm, through the airborne SAR image SIFT matching experiment, from the representative area angle of three different objects category of the selected airborne SAR image, it is proved that the SIFT algorithm can match the stable feature more accurately and has certain robustness. Then, in the airborne SAR image, the airborne SAR image has a certain robustness. On the basis of the matching algorithm flow analysis, the matching experiment combined with the SIFT and the gross error elimination algorithm and the SIFT matching experiment based on the square constraint are carried out on the selected representative regions of the selected airborne SAR images of three different types of ground objects. The results are analyzed and evaluated by the matching points.
The contents and innovations of this paper are as follows:
(1) in this paper, the airborne SAR data of three different types of ground objects are used to do the experiment. In the airborne SAR images containing artificial buildings and lack of natural vegetation and texture information, the SIFT algorithm can effectively extract the stable matching points from different angles, and the correct rate is high, even in the area lacking texture information. In practice, the matching of overlapping areas between SAR images is achieved.
(2) the 2D single stress transformation was introduced to remove the SIFT mismatch point pair as the RANSAC algorithm, and the RANSAC algorithm was used to eliminate the SIFT mismatch point pair. In the RANSAC algorithm based on the 2D single stress transformation, the airborne SAR image mismatch elimination experiment was tested. By analyzing the matching points to the data, it was proved that the method could effectively eliminate the SIFT mismatch. Matching point pairs further improve the accuracy and accuracy of SIFT matching points.
(3) the imaging mechanism, image feature and image model of synthetic aperture radar (SAR) are analyzed. It is stated that SAR image matching must take into account its own image features. In view of the special imaging features of airborne SAR images, the R-D model is proposed, and POS and DEM data are combined to assist image point positioning, and the geometric constraints of airborne SAR images are added to SIFT. In the algorithm matching, the matching points of the SAR image to be matched are predicted by the object square constraint, and the matching search window is set up to predict the matching point as the center, and the matching is made by using the SIFT algorithm in this constraint range.
(4) the SIFT matching experiment based on geometric constraint, compared with the matching algorithm combined with SIFT and gross error elimination, greatly reduces the mismatch point pair and further improves the accuracy and accuracy of the matching point pair; based on the SIFT matching of the object geometric constraint and constrains the search matching range of the matched SAR image, then the matching range is improved. Matching efficiency. Experimental results of SIFT matching based on object geometry constraint show that the method is a very effective algorithm for airborne SAR image matching.

【學(xué)位授予單位】：山東農(nóng)業(yè)大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2013
【分類號(hào)】：P237

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 王宇;王涌天;劉越;;基于SIFT和小波變換的圖像拼接算法[J];北京理工大學(xué)學(xué)報(bào);2009年05期

2 陳鷹,葉勤,鐘志勇;基于小波變換的雷達(dá)與光學(xué)影像匹配算法研究[J];測(cè)繪學(xué)報(bào);2000年03期

3 黃勇,王建國(guó),黃順吉;一種SAR圖像的自動(dòng)匹配算法及實(shí)現(xiàn)[J];電子與信息學(xué)報(bào);2005年01期

4 張輝;王建國(guó);;基于目標(biāo)檢測(cè)的SAR圖像匹配算法[J];電子與信息學(xué)報(bào);2008年12期

5 王山虎;尤紅建;付琨;;基于大尺度雙邊SIFT的SAR圖像同名點(diǎn)自動(dòng)提取方法[J];電子與信息學(xué)報(bào);2012年02期

6 劉向增;田錚;溫金環(huán);武建明;張朝陽(yáng);;基于仿射不變SIFT特征的SAR圖像配準(zhǔn)[J];光電工程;2010年11期

7 孫艷麗;李建海;周偉;劉佳琪;;SIFT在高分辨率SAR圖像自動(dòng)配準(zhǔn)中的性能分析[J];電子設(shè)計(jì)工程;2011年07期

8 曲天偉;安波;;二維投影變換模型的單應(yīng)矩陣表示[J];信息技術(shù);2008年03期

9 劉小軍;楊杰;孫堅(jiān)偉;劉志;;基于SIFT的圖像配準(zhǔn)方法[J];紅外與激光工程;2008年01期

10 耿則勛,李勤爽;影像匹配中金字塔數(shù)據(jù)結(jié)構(gòu)比較與快速算法[J];測(cè)繪學(xué)院學(xué)報(bào);2000年01期

本文編號(hào)：1894388